The Vault

Power at the edge: Processing and storage move from the central core to the network edge
White Paper / Feb 2017 / 5g, mobile edge computing

MEC and other edge computing initiatives address the need to place processing and storage where appropriate, whether a central location or the network’s edge, depending on factors such as applications, traffic type, network conditions, subscriber profile, and operator’s preference.  In this InterDigital sponsored report, RCRWireless explores the evolution of the edge’s role in fixed and mobile networks and how it may impact network optimization, value-chain roles and relationships, business models, usage models and, ultimately, the subscriber experience.

We found some problems with the form:
To view this content please fill out the form below to tell us about you.
Privacy Policy
REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |1| In collaboration with REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |2| Table of contents I. Report: Power at the edge. MEC, edge computing, and the prominence of location 3 What: Basics 4 Why: Drivers 8 Which: Initiatives 14 Where: Topologies 16 Who: Business models 22 When: Timeline 24 Implications 26 II. Vendor profiles and interviews 27 ADLINK Technology 28 Advantech 36 Artesyn 43 Intel 50 Qwilt 64 Vasona Networks 71 III. Service provider interviews 78 BT 79 Verizon Wireless 85 Glossary 90 References 92 Further resources 94 Watch the video of the interviews REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |3| I. Report: Power at the edge. MEC, edge computing, and the prominence of location REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |4| What: Basics Location, location, location. Multiple-access Edge Computing (MEC) and edge computing in general are gaining acceptance in both fixed and mobile networks as we increasingly realize the power of location in wireless networks ? and especially in virtualized networks. This does not mean the centralized cloud or the big data centers hosting the network core will go away anytime soon. But a rebalancing act is definitely due. In recent years, there has been a strong push to move everything to a centralized cloud, enabled by virtualization and driven by the need to cut costs, reduce the time to market for new services, and increase flexibility. In the process, we lost sight of how important the location of functionality is to performance, efficient use of network resources and subscriber experience. Physical distance inevitably increases latency. Central processing and storage limit the ability to optimize RAN utilization. A fully centralized network may be easier and cheaper to run, but it does not always keep subscribers happy. MEC and other edge computing initiatives address the need to place processing and storage where appropriate, whether a central location or the network?s edge, depending on factors such as applications, traffic type, network conditions, subscriber profile, and operator?s preference. Virtualization, initially used as the basis for moving to the centralized cloud, is even more foundational in enabling hybrid models, because it gives service providers the flexibility to choose location, hardware and software independently to optimize end-to-end network performance and QoE. Both operators and vendors agree that we need to keep a healthy balance between what remains centralized and what gets distributed to the edge. The same applies to the RAN: in some places a centralized ? C-RAN or vRAN ? approach makes sense; in others the traditional distributed model works just fine. In this perspective, virtualization, MEC and 5G, in different but complementary ways, free both fixed and mobile networks from the constraints of a centralized architecture and topology. The new networks can adapt to and accommodate new applications and functions, and can optimize their REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |5| performance. They replace legacy networks in which applications and functions had to painstakingly overlay an existing, rigid architecture. In this report, we explore the evolution of the edge?s role in fixed and mobile networks and how it may impact network optimization, value-chain roles and relationships, business models, usage models and, ultimately, the subscriber experience. Terminology: MEC or edge computing? In this report, we use the term ?edge computing? to refer to the processing, storage and network optimization at the edge of both fixed and mobile networks, that is independent (and agnostic) of the access technology. Specific implementations of edge computing can, however, play a role in optimizing the utilization of access resources. MEC is an example of that. Mobile edge computing refers specifically to mobile networks. Because networks increasingly include fixed and mobile components and they integrate them more tightly than in the past, the distinction between edge computing and mobile edge computing is narrowing. It may disappear altogether with the convergence of fixed and mobile networks. MEC is a specific approach to edge computing that is primarily intended for mobile operators, or, more generally, service providers that have a core network on which MEC can be overlaid. ?Edge? is the term that resists a simple definition. In the context of edge computing, the edge could be in the RAN or the customer?s premises, or it could be an aggregation point in a more centralized location. As networks evolve and become virtualized, the opposition of central core and edge is likely to disappear, to be replaced by multiple locations that may be appropriate to host a given application or function. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |6| Do we still need edge computing in virtualized 5G networks? How does edge computing ? and specifically MEC ? relate to the other main technological innovations in wireless, namely virtualization and 5G? Strictly speaking, MEC does not depend on them, and MEC deployments can start ahead of 5G or full virtualization. However, they share a common direction toward networks that are more flexible and less homogeneous, in turn dictated by the need to increase network efficiency, capacity and cost. As they all move toward the same goals in parallel, they enable and reinforce each other, because they work within different domains in the end-to-end mobile network. 5G, C-RAN and vRAN improve performance mostly through an evolution of the RAN ? e.g., through spectrum utilization, wireless interface, architecture. NFV and SDN work at the function level to optimize processing within the core. MEC is more narrowly focused to enable operators to manage applications and end-to-end traffic at the application level. MEC, 5G and virtualization are not alternative solutions among which operators will choose. We will still need edge computing when 5G arrives. Edge computing lowers latency ahead of 5G, but when 5G arrives, it will need edge computing to lower latency further and meet the 5G requirements. Similarly, RAN virtualization facilitates the rollout of edge computing, because the centralized BBU location where the baseband processing is concentrated provides a good integration point where some edge computing functionality can be located. In a small-cell C-RAN deployment in a retail center, for instance, the BBU location can also host the MEC server that manages location-based applications, which may be available to visitors both over the cellular network(s) and Wi-Fi networks. By combining C-RAN or vRAN deployments with MEC deployments, operators can also improve the business case, because the incremental cost of adding MEC to a new deployment is substantially lower than that of rolling out MEC over the existing infrastructure, as Mansoor Hanif at BT notes in the interview in this report. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |7| What is MEC? No longer mobile edge computing! Among the multiple initiatives to enable or facilitate edge computing and, more specifically, mobile edge computing, ETSI MEC occupies a central role, because it provides a framework to shift processing, storage and control to the edge that is integrated within existing fixed and mobile networks. MEC was created to address mobile operators? need to move the processing and storage of some services and applications to the edge, and to optimize mobile network performance and resource utilization in real time. MEC standardization work started in 2014 at ETSI, with a seminal white paper authored by Huawei, IBM, Intel, Nokia, NTT DOCOMO, and Vodafone. The list of active participants has since grown to include more vendors and operators. The focus initially was on mobile networks, but it now access-technology agnostic and it has extended to fixed networks, to reflect the tighter integration of mobile and fixed networks, which often host the same services and applications. The MEC acronym no longer refers to ?Mobile Edge Computing? and instead stands for ?Multiple-access Edge Computing.? Although the change was prompted by ETSI?s internal process requirements, it was welcome change as it expands the reach and potential of MEC in today?s rapidly evolving wireless networks, which include a more varied set of access technologies and spectrum regulatory frameworks. The extension to non-cellular technologies means that Wi-Fi is now included within MEC?s scope. This is a welcome addition that reflects the fact Wi-Fi in most markets accounts for the majority of traffic to mobile devices. The report will delve into these, but three points are worth emphasizing: ? The main driver of MEC adoption is QoE. With MEC, operators move hardware that traditionally had been located in centralized data centers toward the edge and, more importantly, closer to the users. The two primary advantages of this are a reduction in latency and more efficient utilization of the available capacity ? both of which improve QoE. ? Moving content to the edge can have multiple advantages ? e.g., lower latency, location-aware services, flexible service creation, security, and reduction in backhaul traffic ? but only if the right content is stored locally and the edge location is appropriately chosen. If these conditions are not met, MEC may increase costs and complexity, at a price that is too high to justify the enhancement in QoE. ? MEC emerged within a growing ecosystem that is converging toward a new approach to network design and operations, which we refer to as ?pervasive networks? [25], as opposed to the legacy ?atomic networks.? In this new environment, networks are virtualized, use open source software, and rely on APIs for application development. Technological advances are driven by multiple open initiatives, operators and vendors, reducing the impact of proprietary solutions and even of established standards bodies such as 3GPP. From the hardware side, MEC and edge computing introduce the need for modular solutions that support multiple form factors to enable deployments in a more varied set of environments. Features ?On premises ?Low latency ?Proximity ?Location awareness ?Network context information Approach ?Virtualized network ?Standard-based ?Open source ?APIs ?Multivendor ?Modular ?Convergence of multiple initiatives Goals ?Better QoE ?Lower latency ?RAN and content optimization ?More efficient use of resources ?Central cloud offload ?New services REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |8| Why: Drivers Edge computing?s appeal comes from a growing realization that centralized topologies are not sufficient to serve current and forecasted traffic loads with the QoE that both operators and subscribers expect. The fundamental driver for edge computing is the continued growth in subscriber traffic and, especially, in real-time traffic, such as video and interactive applications such as games. In the future, augmented/virtual reality traffic will add pressure on operators to continue to increase capacity and, just as importantly, reduce latency. High latency ? even in a high-capacity network ? will cripple QoE. For subscribers, limited capacity and high latency may have the same effect and look indistinguishable. For an operator, adding capacity without lowering latency can be an expensive mistake. Even with 5G?s promised reduction in latency, edge computing will be useful in reducing the latency introduced by the backhaul. While backhaul technologies add different levels of latency, they all inevitably contribute to it as a function of physical distance. The concurrent growth in traffic load ? increasing capacity requirements ? and of real-time traffic ? increasing latency requirements ? drive the need for processing, storage, and control for selected applications to be moved to the edge. At the same time, operators face a challenging situation because they have to meet high QoE expectations in a cost-effective way, in an environment where subscriber revenues are flat in most markets. This means that they need to improve performance within the current spending levels. To do this, they have to maximize the utilization of existing resources before they launch into a RAN expansion. MEC and edge computing can play a major role in doing this, because local control and processing enable a finer-grain real-time optimization of RAN transmission. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |9| Growth in video traffic increases importance of latency Video traffic keeps on growing, and at a faster pace than other types of traffic. According to Cisco?s VNI, mobile video accounts for 60% of traffic today, and that number will likely be 78% by 2021. Ericsson?s estimates are in the same range: 50% in 2016, growing to 75% in 2022. The 2016?2022 CAGR for video is 50%, compared to 23% for web browsing and 39% for social networking. The growth in video traffic is hardly surprising, given the availability of more video content and its higher-quality. Also, video is no longer confined to video apps: it has become an integral component of social networking and communication apps. In the process, real-time video calls have finally gained the social acceptance that, for a long time, was missing and slowed down the adoption of video communication as an alternative (or complement) to voice or text. The increased use of video in its multiple forms ? e.g., downloaded, streamed, uploaded, interactive ? reinforces the role of latency in determining overall subscriber experience (and the attendant churn rates). For this reason, video has, since the beginning, been one the main drivers for edge computing. Not only is video traffic growing it is often location-specific and concentrated in specific locations and times that are the backdrop for perfect use cases to justify moving processing and storage to the edge. Edge computing cannot lower the latency in the RAN, but it can reduce the end- to-end latency: video can be cached at the edge, or the processing (e.g., with ABR) that optimizes video traffic can be done at the edge. Both approaches can be combined to reduce the traffic between the centralized core and the edge. Video is an instance, however, in which it is crucial to pick the edge?s location within the network carefully to avoid unnecessary investment and complexity. For instance, caching at the edge works because subscribers at a given location and time tend to watch a remarkably consistent and narrow set of content. But if the MEC server is co-located with a small cell, there are too few subscribers within the cell footprint to justify caching, even if the cost of storage were not an issue. At the same time, the more remote the integration point is, the lower the impact on latency and hence on QoE. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |10| We want it all, we want it now, we want it here: immediacy and time/location awareness The use cases for mobile computing go beyond the need for low latency in video and real-time traffic. Immediacy is another performance yardstick that subscribers care about and expect in all applications, not only real-time ones. Immediacy can be quantified as the time it takes for an application or service to launch, or for content to appear on the device or register as being sent. Edge computing can bolster immediacy, because content no longer needs to be downloaded from a remote data center, or to be sent to the data center for processing and then back to the subscriber. Managing applications and traffic types in real time at the edge enables operators to control the effects of congestion on time-to-content and to provide the desired level of immediacy. In enhancing QoE, the relevance of and application or content is complementary to immediacy ? and relevance is often tied to time of day and the subscriber?s location. Video content popularity peaks and declines very quickly. An offer for a discount on a restaurant meal works best if you are near the restaurant and hungry. Mobile operators already provide time- and location-aware content, but edge computing can make that content more closely tied to the location and hence more relevant to subscribers. In addition, time- and location-aware content appeals to content providers and venue owners that may be the source of the content and applications hosted on the edge servers. Stadiums are a good example of a location with high traffic loads that are mostly location/time dependent and have a high perceived value to the spectators. Videos from the game, content associated with teams, or ads for services available at the stadium generate a massive amount of content ? and to a large extent, it is also consumed locally. Storing and processing that locally, at the network edge, will improve QoE for those attending the game. And by offloading that work from the rest of the network, it will improve QoE for other subscribers, too. MEC servers in stadiums may be used to provide location-based services and content not only through the network of a specific operator, but also 2017 Super Bowl: AT&T stats $40 million investment in network expansion 9.8 terabytes of data during the game, equivalent to 28 million selfies 88% increase in traffic over last year 148% up from average pro football game traffic 59.9 terabytes in the Houston area over the weekend Source: AT&T REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |11| across operators and the local, possibly stadium- operated Wi-Fi network. Initially these services may include location-based retail and advertising, services for the public, event-specific applications, surveillance, and stadium operations and services. Farther in the future, they may expand to AR/VR services and IoT applications. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |12| From MEC proofs of concept to use cases If you have heard of only one use case for MEC, it almost certainly involves video. However, the scope of edge computing is much wider. It is driven by the diverse needs of multiple stakeholders: ? Keep latency low for real-time traffic ? video, but also voice for enterprise environments ? Offer applications and content that target visitors in a venue and may be finely tuned to their location (e.g., proximity to a store) or timing (e.g., off- peak hours) ? Keep local content within the premises; avoid sending content that may be generated and accessed locally, or that is location-specific (e.g., enterprise data) to the centralized core, and back to the RAN to be delivered to the subscriber ? Share local content among available wireless networks, including Wi-Fi and multiple mobile service providers ? Provide an extra layer of security by keeping local content within the enterprise premises ? Optimize content delivery based on real-time RAN conditions, as well as other factors, such as subscriber devices, content type, application requirements ETSI MEC PoCs show areas of early interest from participant operators and vendors. As MEC gets deployed, the list of use cases will grow. The table on the next page lists some that have attracted attention to date. MEC stakeholders: who gains what? Mobile operators and other service providers: provide better QoE, better resource utilization, offload from the centralized core Venue owners: offer venue-based services to visitors, enrich their experience and encourage them to share it Enterprise, IoT: develop and support enterprise-specific applications and services, have fast and secure access to enterprise data and applications over multiple networks, and support IoT applications within the enterprise Content providers, OTTs, application developers: optimize content delivery and application access in real time, adapting content to RAN conditions and subscriber requirements Subscribers: better access to applications and content, increased immediacy and relevance of wireless connection, leading to a more positive QoE REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |13| REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |14| Which: Initiatives Edge computing requires more effort and investment than adding a server at the selected edge location. Many moving parts have to come together to enable the stakeholders ? service providers, venue owners, applications/content providers, enterprises, subscribers ? to fully harness the benefits of shifting functionality to the edge. Correspondingly, there are multiple, complementary initiatives converging to create an ecosystem that will support distributed network models end to end. Each stakeholder will likely find only a few of the initiatives relevant, and may feel overwhelmed by the apparent competition among them. For vendors, the best bet is often to be active in multiple initiatives to prepare themselves to participate successfully in the nascent ecosystem. Among the factors that converge in the creation of the edge-computing ecosystem are these: ? Edge hardware. It may have to be installed outdoors or in locations with space, security or environmental constraints that differ from those of a data center or central office. ? Mobile devices. New device types will populate the network to support IoT applications, and many of them are likely to benefit from edge processing. ? Services and applications. These may be targeted directly at subscribers, enterprise workers, or visitors to a venue, or may be used for IoT; they may be managed by different entities ? mobile operators, other service providers, venue owners, enterprise, OTTs or content owners. ? Integration in the end-to-end network. Edge functionality has to be tightly integrated with the RAN and the centralized core. Managing potentially dynamic edge locations (i.e., the edge location changes in real time depending on network conditions) requires orchestration capabilities in the core network. ? Integration across networks. The operators of multiple networks, possibly owned and managed by different parties, may want to share edge processing and storage resources in the same location. ? Application developers. They need to optimize their apps or develop new ones to work in an edge environment. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |15| REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |16| Where: Topologies We have talked so far about moving processing, storage and control to the edge, as if it were clear what we mean by the edge. But it is far from obvious where the edge is ? or, more accurately, where potential edge locations are, and which one or ones a service provider should select. This is a crucial question ? perhaps the most important one ? to ensure that edge functionality brings both performance and financial benefits to the service provider and the other edge stakeholders. If the edge location is too far out, too close to the subscriber, edge computing may become overly expensive and complex. If the edge location is too close to the centralized core, the benefits of edge computing dissipate, with a more complex network topology but no significant improvement in performance. And is there a single edge? Not only may different service providers pick different edge locations for their networks or specific locations in their networks; it may also make sense to have multiple edges in a given location, depending on the applications. Location-based content and applications are most likely to be hosted in an aggregation point that reaches all the infrastructure that covers the venue. An enterprise deployment may be housed in a location that covers all the enterprise?s buildings or just a subset of them. For applications that require video caching, service providers have more flexibility in choosing the edge location. They may want to see what their subscriber usage patterns are, and pick an edge location where they can maximize the caching contribution. And is the edge a fixed location? It does not have to be, although initially it is likely be. For many applications ? e.g., location-based and enterprise applications ? an edge location that does not change though time may be desirable. But for locations with highly fluctuating network loads or for applications with uneven temporal and spatial distribution, a moving edge that shifts depending of real-time network conditions is possible in a virtualized environment and can maximize the cost/performance benefits of edge computing. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |17| MEC architecture To address the issue of where the edge is or could be, it is useful to review the MEC architecture proposed by ETSI. Other edge computing initiatives rely on the MEC architecture or use their own edge server. Traffic to and from UEs that involves applications, services or content hosted in the edge server is directed to the server; the rest of the traffic is routed to the centralized core as usual. The MEC server or edge host uses a virtualized platform to host applications It can interface with cellular (3GPP) networks, as well as other available networks, including Wi-Fi networks. Application developers have access to API to get their applications hosted in the edge server. The MEC server provides the processing and storage capability to support the hosted apps. Storage may be used for caching frequently accessed content, or for local breakout (to keep local content within the MEC footprint), and avoid using backhaul resources to transmit the content back and forth. Local processing enables applications to optimize their performance. Another important and innovative element in the MEC server is the addition of user- and network- information services. These provide the foundation for optimizing end-to-end network performance. MEC services for network optimization Radio network information service (RNIS) ? Up-to-date radio network conditions ? Measurements and statistical information related to the user plane ? Information about the UEs served by the radio node(s) associated with the host (e.g., UE context and radio access bearers) ? Changes in UE information Location information service ? Location information: cell ID, geolocation, etc. ? Location of specific or all UEs served by the radio nodes associated with the ME host ? Location of a category of UEs (optional) ? Location of all radio nodes associated with ME host Bandwidth manager service ? Allocation of bandwidth to ME applications ? Prioritization of certain traffic routed to ME applications Source: ETSI REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |18| The MEC server is the bridge between what happens in the RAN and the UE and what happens with the applications and content. It enables network operators, application and content providers, and others that may play a role in serving subscribers to manage traffic in real time on the basis of factors such as application/content requirements, network conditions, and policy ? to optimize utilization of the available network resources. Network operators can use this information to manage core and RAN resource allocation, but they can also share this information with content owners, venue owners and application providers to coordinate traffic management with them. Where is the edge? Location tradeoffs If cost, complexity and space were not constraining factors, moving processing and storage to the far edge would generally be the best way to improve QoE. But they do matter, and so the choice of edge location does not rest solely on performance, but on the evaluation of tradeoffs among multiple factors. The criteria for such evaluation will vary among operators, but there are some high- level considerations that apply across networks. As the chosen edge location gets closer to the UE: ? The latency gets lower, and this improves performance on applications that are sensitive to it, such as video calls, voice or gaming, leading to a higher QoE. ? Hardware has less storage capacity, limiting the amount of content that can be stored at the edge. ? Processing power becomes more limited, and hence it may not be efficient to run some applications from the edge. ? End-to-end network complexity increases, because network operators have to deploy, integrate and manage more hardware, at a higher number of locations. ? Resources may be needlessly duplicated if applications could be efficiently run from a centralized location. The opposite side of the equation holds when the chosen edge location moves toward the centralized cloud (i.e., higher latency, more storage capacity, more processing power, less complexity, more efficient use of resources). Another key consideration is the footprint ? it gets smaller as the edge host moves away from the centralized cloud ? because this is the area over which the RAN optimization and the mobility management are scaled. When the edge host covers a small area, edge traffic optimization is confined to this area. So, two MEC servers can both optimize their respective footprints, but as two separate REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |19| zones. If a single MEC server covers both footprints, optimization can be coordinated across the entire area. Coordination across a wider area may make optimization more effective, but at the same time the greater distance to the RAN may limit the granularity of the optimization capabilities. The footprint of the edge host is also determined the type support for mobile access in edge-based applications and services. Within the footprint, subscribers? experience is preserved as they move from one cell to the next within the same network. If the edge host covers multiple networks, application access can be preserved across networks. As the subscriber moves away from the footprint to an area that has no edge host or a different one, access to the application has to be managed to ensure a smooth transition. When the two footprints do not have the same edge capabilities, the subscriber may experience a discontinuity in the quality of the access to the edge application. Obviously, if the application is supported only at the edge (e.g., an enterprise application or a location-based application in a mall), subscribers lose connectivity to the application as they move out of the footprint. In many cases this is a desired outcome (e.g., the enterprise may want its services available only within its campus), but it is something that has to be kept in mind when selecting the edge host location and how to manage mobility across footprints. Factors that play a role in the selection of the location of the edge server include: ? RAN resources. The edge server capabilities must be sufficient to serve the covered footprint. If they exceed the RAN capabilities, the investment in edge computing is wasteful. ? Backhaul resources. Edge computing may address capacity and/or latency limitations in the backhaul and prevent the backhaul from becoming a performance bottleneck. ? Applications. Latency, processing and storage requirements that affect edge location vary across applications, so the ideal edge server location varies by application. ? Subscriber/client expectations, and venue-owner preferences and requirements. The expectation for service performance and QoE may be different for enterprise employees, mall visitors, or IoT sensors. ? Operator policy and preferences. Operators may want to position themselves in the market in a specific way ? e.g., provide a higher-quality service for a specific enterprise client or venue owner. ? Content provider preference. If the content provider pays for the edge infrastructure, it will want to choose the location of edge hosts, because this will allow the provider to maximize its return on investment. In addition, real-time optimization enables operators to shift the location of the edge dynamically depending on factors such as demand for an application, RAN conditions, concentration of subscribers or power considerations. A virtualized platform will make it possible to instantiate applications at different edge locations as operators expand their edge computing and virtualization capabilities. Where is the edge? Location options vCPE, home/office GW, HeNB RAN Wi-Fi access points Corporate/venue/factory gateway Cable boxes Outdoor furniture C-RAN BBU pool Central offices Cable/operator gateway RNC Other aggregation points REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |20| Optimizing end-to-end network resource utilization from the edge Improving performance is an obvious goal for mobile operators. But when it comes to ensuring cost effectiveness and profitability, network resource utilization is the goal that should take center stage. It is a measure of how much value operators can squeeze out of the network infrastructure they have ? which in turn is a measure of subscribers? happiness with the service. In today?s networks, the prevailing approach is to maximize performance, given the financial resources available. This typically means increasing RAN capacity using a brute-force approach ? i.e., deploying the latest technological tools that increase throughput. But far less effort is put into optimizing the use of the capacity available. It is like buying a fast car without having roads that allow you to drive fast. There are many ways to optimize network resource utilization and, under intense pressure to improve performance without increasing costs, mobile operators have started to work toward this optimization goal. MEC and edge computing in general are geared to achieving exactly that by changing the processing, storage and control in the network. The impact on QoE from moving processing and storage to the edge is easy to grasp, even though it is not trivial to quantify over a network because it depends on multiple environmental factors that are variable. Other things being equal, though, moving processing and storage to the edge improves latency, immediacy and QoE. The new control features at the edge introduce a new type of optimization, one that works in real time, leveraging information about network conditions to optimize end-to-end network performance instead of optimizing the performance of individual network elements. Edge computing is not required for this type of optimization ? it can be implemented in current 4G networks ? but MEC servers are well suited to gathering information from the RAN, processing it and forwarding the results to the centralized core or to content or application providers. ETSI specifications define services (see table above) that collect information that can be used in multiple ways to optimize the utilization of network resources and QoE. Some optimization is best done, either remotely or at the edge, by the entity that controls the applications and content ?the mobile operator, an OTT or a content owner ? because that entity has direct control and better knowledge of content and applications, as well as better access to them. And it often wants to retain some degree of control to ensure it can create the subscriber experience it aims for. An example is that, increasingly, the content served to subscribers is encrypted; operators do not know the content type, much less have the flexibility to optimize it, but the content owners do. Throughput guidance is an optimization tool that is being developed to address the mobile operators? need to adapt content to real-time RAN conditions and to be able to do so in collaboration with third parties. It uses data about network conditions ? especially RAN conditions and RAN load ? to generate advice for content and application providers on how to manage traffic exchanged with the subscriber. When the network has sufficient capacity, the providers can share content at the highest quality available. When the network is capacity constrained or congested, the content and TCP transmission can be adapted to provide subscribers the best experience possible given the real-time availability of network resources. Tools like throughput guidance allow operators to tell third parties what they can do to optimize content delivery, without the operators having to access it directly. For it to work, however, operators and third parties have to tightly coordinate network optimization. In some cases, this will require getting past a history of tension and competitiveness between mobile operators and content/application providers. The situation is rapidly changing, however, as REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |21| both camps realize they need to work together to provide outstanding QoE. And both sides might benefit further if throughput guidance and other optimization tools are the catalysts that facilitate tighter relationships among them. Throughput guidance: results from Google and Nokia trial Network metrics improvement TCP retransmissions 30-45% TCP round-trip time 55-70% Mean client throughput 20-35% TCP packet loss 35-50% Application metrics Click-to-play time reduction 5-20% Average video resolution improvement 5-20% Video format change frequency reduction 10-25% Source: Nokia, Google REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |22| Who: Business models Mobile operators will continue to have a leading role in planning and funding MEC deployments, because they are integral parts of their network infrastructure. Other edge computing hosts, partially or fully independent of mobile operators, are likely to be deployed, paid for, and operated by the enterprise or venue owner, even if some of the applications supported can be hosted by the mobile network. Even in the MEC space, however, new business models may arise that have a more direct and active role for venue owners and the enterprise, on the periphery side, and for content/application providers, on the cloud side. They both stand to benefit directly from the MEC infrastructure ? and, in some cases, more than mobile operators do. For example, a MEC server that supports industrial IoT applications in a warehouse may be more valuable to the enterprise than to the mobile operator. The enterprise may see in it a compelling business case, while a mobile operator might struggle to see a positive ROI or might not be able to assess the revenue potential because it is dependent on enterprise-specific applications that it is not familiar with. Similarly, a content or application provider may be willing to locate some of the infrastructure it needs at the edge of the network when that is more effective ? and potentially more cost effective ? than a remote cloud location. And it is not only companies like Google or Facebook that may be interested in having a presence at the edge. Smaller companies may also be willing to locate processing and storage functionality at the edge in a virtualized environment, where they do not need to own a host server but might pay only for the services they need. In this model, the mobile operator may deploy and pay for the initial edge hardware, but then it can monetize the investment by renting access to it to a third party. An arrangement of this type can be mutually beneficial, especially when accompanied by joint efforts to optimize network performance ? for instance, with tools like throughput guidance. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |23| Edge computing stakeholder Why pay for edge computing? Mobile operators ? Better QoE. Churn reduction, lower customer support costs. ? Increase in network resource utilization. More value extracted from existing infrastructure, better end-to-end network TCO, need for capacity expansion. ? Location-based services. Monetization of services to subscribers (possibly), to the enterprise, to public venues, and for IoT services. ? Offload centralized core. Cost-effective improvement of both QoE and resource utilization. When the centralized core requires additional processing/storage capacity, operators may decide to deploy it surgically at the edge where needed. Other service providers, including IoT providers ? Location-based services. Service providers, such as DAS neutral hosts, cable operators, wholesale service providers or MVNOs, may be willing to install their own edge infrastructure to provide services specifically targeted to a location. They might use a local access network they own or manage, or lease capacity from the RAN local network operators. They may monetize such services to venue owners or other parties with a presence in a venue (e.g., retail) or with IoT applications. Public venue owners ? Location-based services. Offered as an amenity to guests (e.g., stadium, hospital) or as a service (e.g., city, college); to advertise to visitors; to support needs of tenants (e.g., stores in a mall); to support their own operations. Services can be made available through Wi-Fi or MulteFire venue-owned networks, DAS networks or small cells. Enterprise ? Local breakout. Keep enterprise data and applications local, provide enterprise voice services. ? Security. Keep enterprise data and traffic local to the enterprise. ? Enterprise services. Develop and possibly mange enterprise-based services and applications. Services can be made available through Wi-Fi or MulteFire venue-owned networks, DAS networks or small cells Content and application providers ? Shift of processing and storage to the edge. Improve QoE, better control delivery of service, coordinate real-time traffic optimization with operators. Content and application providers may own their edge infrastructure, but leasing resources from the operator?s virtualized edge servers is an approach more likely to be accepted by both operators and third-party providers. Residential and small businesses subscribers ? Home/small business gateway. Residential and small-business customers may invest in an edge host that supports services and hosts content used by the people within the premises and shared over the Wi-Fi and cellular networks. Service providers may subsidize the edge host as a subscriber-retention feature. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |24| When: Timeline In many ways, edge computing is nothing new. There have been edge computing solutions all along to serve niche markets or to address specific performance and optimization challenges in mobile networks. What is different today is that network virtualization offers a framework to expand edge computing capabilities, adding scalability, reliability, flexibility, cost effectiveness. This will take edge computing to the mainstream and enable operators to reap benefits in terms of improved QoE and resource utilization. The ETSI MEC standardization work creates the foundation for edge computing deployments in mobile and, increasingly, fixed networks. During the first term (2015-2017), ETSI ISG completed the groundwork, released the basic specifications, and encouraged the creation of the ecosystem. More work is needed during the second terms (2017-2018) not only to expand beyond mobile networks, but also to strengthen the links with other edge computing initiatives while avoiding the risk of fragmentation of efforts. Beyond standardization and industry collaborative initiatives, there is a need to explore different business and deployment models, and revisit the role that stakeholders ? e.g., venue owners, enterprises, content and application providers ? will have in deploying, managing, and funding edge computing deployments. The business case also needs to be assessed to understand where and when edge computing provides a better return that the centralized cloud. To assess the business case for edge computing we need to go beyond the standard ROI model. Improvements in QoE or resource utilization are highly valuable, but notoriously difficult to quantify, because they involve end-to-end network improvements. A traditional financial model that looks at a solution that delivers a well-contained benefit is inadequate for edge computing, as it does not adequately capture the costs that it requires and the value it brings. The time to commercialization can be fast as edge computing can be introduced without waiting for full network virtualization or 5G. In practice, however, it will take a couple of years before commercial launches, as vendors and operators complete their trials to learn what is the most efficient way to balance centralized versus edge processing and storage. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |25| ETSI MEC second term objectives ? Support 3GPP access technologies (Wi-Fi and fixed) ? Extend the virtualization support types, to render the environment as attractive as possible for third-party players ? Study possible charging models which may be applicable to MEC ? Fill gaps relating to lawful interception ? Develop testing specifications and test methodologies ? Coordinate plug fests ? Coordinate experimentation and showcasing of MEC solutions ? Expedite the development of innovative applications ? Ensure a low entry barrier for application developers ? Disseminate the results of the work ? Strengthen the collaboration with other organizations ? Study new use cases ? Enable MEC deployments in NFV environments Source: ETSI REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |26| Implications Mobile edge computing takes us beyond the centralized cloud, to hybrid virtualized model, which combine centralized and distributed processing, storage and control Operators can leverage network flexibility to find the best edge location to maximize QoE and optimize network resource utilization, the main drivers for edge computing MEC is not for mobile networks only. Other fixed networks including Wi-Fi can use the MEC framework and share it with mobile networks There are multiple network edge locations where it makes sense to deploy MEC servers or other edge hosts. Evaluating the tradeoffs that these locations offer is crucial for successful edge computing deployements New business models will accelerate move to the edge, with an increased role of venue owners, enteprises, and application and content providers. Application and content optimization at the edge encourages a tighter cooperation of mobile operators with application and content providers REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |27| II. Vendor profiles and interviews REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |28| Profile ADLINK Technology Since 1995, ADLINK has provided hardware and software solutions for a variety of indoor and outdoor environments in multiple form factors. The core of ADLINK?s business is serving the market for industrial and telecommunications applications, which is becoming increasingly interconnected with the internet of things. ADLINK provides embedded solutions ? such as network appliances, hardened outdoor products, vCPEs, and industrial IoT equipment ? ranging from modules to network appliances and servers. The products address multiple markets, including industrial, military, health, transportation, utilities and telecommunications verticals. ADLINK also provides measurement and automation products for industrial systems, machine vision systems, and automated test and measurement equipment. A third market area includes smart displays, and fixed and mobile computing platforms for the same verticals, with a focus on operations in harsh environments. Edge computing is a good fit within the scope of ADLINK business. ADLINK provides hardware solutions that enable the deployment of functions at the edge that support multiple-access edge computing (MEC) and fog, comply with industry standards, and encourage interoperability. In collaboration with ecosystem partners, ADLINK offers pre-validated software and hardware to facilitate and accelerate edge computing deployments. In addition to its data center and central office solutions, ADLINK offers the SETO-1000, an Intel Xeon-based, ruggedized edge server for outdoor use that targets the MEC and fog market. The SETO-1000 uses a 19" chassis, has up to 96 GB of RAM, and can support applications such as augmented reality, video analytics and caching, and distributed content and DNS caching. ADLINK has recently launched the SETO-1000 as part of one of the MEC architecture?s three main elements: ? The MEC hosting infrastructure management system: the SETO-1000, the virtualization layer and the virtualization manager ? The MEC application platform management system: traffic control RAN information services, communications services and service registry ? The application management system: the MEC virtualized machine For next-gen virtualized networks, ADLINK is working on solutions that use ADLINK?s Modular Industrial Cloud Architecture, in which hardware and software are decomposed and are based on open specifications. This architecture operates in real time, because the confluence of virtualization and IoT imposes a resources-on-demand model in which hardware resources must be flexible enough to meet competing needs from multiple applications in real time. Achieving this, ADLINK argues, will take more than modifications to the current hardware; what is needed is a more fundamental redesign of the underlying architecture, in which computing, storage and I/O are allocated to separate functional models during the design and then combined, as needed, in the hardware unit, based on the specific requirements. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |29| ADLINK Technology Network optimization at the edge A conversation with Jeff Sharpe, Senior Product Manager, Network & Communications, ADLINK Technology Monica Paolini: Network virtualization adds flexibility and scalability to networks, and enables operators to use network resources more efficiently. The ability to move network functions to the edge is a crucial element in optimizing network performance. The location of network functions becomes more relevant in a virtualized network. In this conversation, Jeff Sharpe, Senior Product Manager, Network & Communications at ADLINK Technology, shares his views on what is required to move network functionality to the edge and what the benefits are. Jeff, let?s get started with an introduction to your role at ADLINK and how ADLINK entered the edge computing market in telecoms. Jeff Sharpe: I?m the Senior Product Manager of one of our business units, called Network & Communications. My primary focus is long-term strategies for products. ADLINK has been around for over 20 years. We?re leaders in edge compute technologies in multiple technology segments ? industrial IoT, network and communications, telecom industries, even military. We are a hardware provider, but we also supply software services. We have a long history and growing expansion of ecosystem partners, and we can come to the table with our telecom friends. Monica: MEC is part of ADLINK?s identity. You have been doing edge computing for a while. Jeff: Absolutely. With our leadership in industrial IoT and manufacturing, we have been making a lot of the key edge components for years, and we?re now starting to see them in the telecom and mobile markets. Some of our existing partners ? such as Intel, Nokia, Saguna, Tieto, Wind River or PeerApp ? were early players. They participated in some of the first deployments while driving the standards that mobile edge computing started to develop about two to three years ago. We?re way ahead of the curve, and very excited about this technology. Monica: What has ADLINK done so far with its partners? Jeff: Saguna is a great partner of ours. They?ve been around since day one of the ETSI mobile edge compute. We realized that for us to be successful and for our customers to be successful, we need a system that?s not just a piece of hardware you would put at the edge, but that also has carrier-grade capabilities, that fits ETSI standards and is based on open architectures for the RAN, customer premises, outdoor cabinets or the central office. Saguna is a great fit with that. They have great software that we embed within our systems. Tieto, another one of our partners, does a lot of work on virtual BTS, virtual radio access networks. We also supply some of their software to service providers. Another one of our critical partners is Wind River Technologies, who provide Intel-based NFV/SDN/vCPE software called Wind River Titanium Core and Titanium Edge that ADLINK embeds within our systems. This list continues to expand. MECSware is another partner, based in Europe. It?s a branch off of Nokia, and we?re collaborating on some virtual RAN opportunities. And finally, we collaborate with security companies such as Trend Micro, Fortinet, and Checkpoint, and vCPE providers such as netElastic, Wind River, and CertusNet. With our partners, we work on proofs of concept with the service providers. Monica: How is the involvement of mobile operators worldwide progressing? Jeff: What we?re seeing is that a lot of the service providers are in different stages of mobile edge compute in different countries. We see Korea, especially SKT, as being one of leaders, especially with 5G coming out. They want to use more and more MEC technologies to speed up video and analytics, especially. Europe seems to be at the forefront of MEC as they?re including it with their LTE rollout. Proofs of concept started with a focus on network optimization. A lot of the carriers were looking at MEC and asking, ?How can I optimize my network? How can I save money? What?s my total cost of ownership savings that I can have in my network?? Some of the earlier proofs of concept were about video caching, augmented reality, and content REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |30| delivery to improve QoE while reducing impact on the backhaul networks. ?How can I do more at the edge versus in my cloud?? It?s starting to shift. We started in Europe. Now in the US, we?re starting to see more proofs of concept centering on the monetization of the network. How can the service providers generate more money while improving services, especially better quality of experience for their end customers? The proofs of concept are now more into analytics for network grooming for 5G, and providing valuable services to help monetize their networks. Operators have many questions. ?How do I get prepared for 5G, enable augmented reality, virtual CPE, virtual BRAS, which is more of a broadband access service? How can I generate more money? How do I improve quality of experience for my customers?? We?re seeing a change in deployment of applications based on regions. In China and Asia- Pacific countries like Japan, we?re starting to see more operator uptake on services as well. In the rest of Asia and in Oceania, we?re seeing more security, deep packet inspection on how they can reroute traffic, load balancing, and, again, network optimization. It?s an exciting area, because there are so many different applications you can run at the edge. By the way, Monica, I wanted to define the edge. We see the edge as anything that is outside the data center and closer to the customer. From an ADLINK perspective, we see the edge being the customer premises, or the RAN tower itself, or the central office. We see that as being very, very close to the customer, and optimizing the utilization of the operator?s key assets ? that?s the tower, the central offices, and even their customers? premises. Monica: You?re working on multiple fronts. How are you supporting your partners and customers? Jeff: We?re seeing things starting to gear up, mainly from the central office towards the customer premises. We have products that fit in each of those areas. We have an outdoor server that fits directly into the radio tower itself. It sits outside. It?s waterproof and it?s fanless. A primary question is, ?How do I replace a lot of that equipment at my RAN tower and virtualize all that equipment on a Xeon-class server?? That was part of our early proofs of concept with Verizon. We were also embedded in some of the Nokia products and their Liquid server program. It?s exciting. Now we?re starting to see all these other new applications coming in ? video monitoring, video surveillance, and traffic management. All these cool applications are widespread at the edge. We see great potential there. Monica: You mentioned multiple edge locations. Does each operator have a different preference for where it wants the edge to be, or does it depend on the application? Jeff: It?s a little bit of all that. That?s a great question. If you look at virtual CPE as an up-and- coming application, a lot of carriers are saying it looks like it?s going to be a $1.5 billion market by 2019 or 2020. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |31| It could be these little, tiny boxes that you actually put on the customer premises, like a Starbucks, a McDonald?s or a small franchise. Or it could be a larger data-center environment, where the telco would place their equipment in that data center or in a closet, so it needs a bigger space. When you look at things like virtual radio access networks, the operator needs to virtualize a lot of key equipment at the tower. As the network grows, things like load balancing, traffic shaping, content injection, lawful intercept, all of these key programs that use DPI, would use larger servers in the central office. Seeing the whole gamut of infrastructure at the edge, we initially thought that we?d put everything in the cloud and everything would all be perfect. But the latency increases and the backhaul network introduces delays. Also, the cloud may not have the horsepower to host the number of appliances that are going to be attached to the network. That?s where the critical part of computing at the edge comes into play. Monica: What is the security advantage of moving functionality to the edge? Jeff: The key aspect is securing the service provider?s customers. One of the key assets of things like vCPE is offering next-gen firewall protection, DDoS protection, or encryption. There?s value in the ability to do IPsec from the customer all the way to the customer?s application, which could be in their private cloud, and offering as a service the encryption of all the messages from the vCPE end user and equipment all the way to their cloud. Security is, of course, one of the most important aspects. The service provider can offer these security services without the end users having to figure out what type of security they want to use. Again, this comes back to the monetization of the operator?s network and using that entire infrastructure to benefit the operator?s end user. Monica: You said that to make sure the ecosystem is in place, you have to work with multiple partners and customers. At the same time, there are also many initiatives to support edge computing. MEC is maybe the most talked-about one, but there are many other initiatives, too. How do you see them working together? Jeff: We?ve been part of the ETSI MEC, which is the ETSI version of mobile edge compute. With this involvement and through our partners, we?ve enabled the MEC infrastructure for deployment. We?ve also been leaders of another edge technology for IoT called the OpenFog Consortium. Moving into 2017, more committees and sponsored consortiums are coming into play. These include Central Office Re-architected as a Datacenter (CORD), Telecom Infrastructure Project (TIP), Open Compute Project for Carrier Grade systems (OCP-CG), and the Open Edge Compute. New ones are starting almost every week. I think it?s great. We?re setting up these different consortiums, committees and new research groups to help the service providers look at the openness of the software and the hardware. It allows industry leaders like ADLINK to come in, provide context, and drive some of the standards in these different areas, whether it?s IoT at the edge, mobile edge compute, or computing edge, which enables our product development and our partners. But how do we have all these initiatives talk to one another? Although the original intent of having these consortiums is great, it?s also confusing the industries on overlap and messaging. Do I use ETSI MEC? What is the difference between TIP and OCP? Do I use CORD? Do I use Open Edge software? What types of standards do I want to use and how do they all interoperate? These are the same early issues seen in NFV deployments using open source as it overcomplicates the service chaining and deployment. Then you throw in 5G concepts for faster and agile mobile networks. How do all these work together and come together long term? How do I optimize and monetize my network? Time will tell which of these standards work out and which ones service providers want to use. I think we?re also hurting ourselves by adding more and more of these industry consortiums to the mix. It?s just adding to the confusion. I would love to see us all come together, work together. Luckily, ADLINK is involved in a lot of these consortiums, so we can bridge some of the gaps and give our ideas to each one with a single focus. Time will tell. Since we?re at such early stages of edge computing, especially in the mobile area, I think the ETSI MEC will be the forerunner, and then we?ll get into CORD-M, OpenFog Consortium and others down the road and see how they can interact, especially with Open Edge. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |32| The question for the member companies will be how to make software available to suppliers as open source so the operators can use it within their network. Monica: Do you think there is a risk of fragmentation or duplication of efforts with all these initiatives? Jeff: Absolutely; I?m seeing it myself. I?m doing the stuff with CORD. I?m doing it for ETSI MEC and Fog Consortium. Our ecosystem partners are focused on each of these different areas too. Individually, they may not be aware of what ETSI MEC is doing, and what CORD is doing, and what Open Edge is doing, and OpenFog Consortium. Which leads to too many ingredients for a simple deployment of the edge. ADLINK is trying to simplify and deliver a standards-based, open architecture for our customers. A great thing happened last year during the MEC World Congress in Munich. ETSI MEC invited the OpenFog Consortium for a day-long session, enabling collaboration between OpenFog Consortium and ETSI MEC to go through the intent of how to work together and what needs to occur to co-partner ETSI MEC. Luckily, we?ve been part of both committees, and it was great seeing that first partnership between two consortiums and two committees starting to work together. Hopefully, that will expand into CORD, TIP, OCP, 3GPP and ETSI NFV. Monica: What does ADLINK do to help its customers select the most relevant initiatives? Jeff: Based on my experience over the past few years in the MEC industry, as we?re starting to see more deployments and more proofs of concept, the operators and the service providers do need help. Mobile edge is fairly new in the US and Europe. Operators are trying to figure out what they want to do with edge computing. How do they implement it? What is the easiest way to take your technology and your partner?s technology, and implement it into their network? They need help and guidance. ADLINK comes into the customer?s situation wearing a solution hat. We have a series of products that fit all the different hardware needs and routing software that are part of mobile edge compute. We also bring in our reference partners ? partners that have been validated on our hardware, or whose software we embed within our hardware for a full PaaS (platform as a service). This way we can de-risk a lot of the proofs of concept and some of the concerns that carriers may have. As I said earlier, operators are really looking for partners and help. Moving away from the typical procurement model, operators are looking for risk sharing and partners to assist them with their decisions and deployments. For example, an operator may want the ability to do 5G analytics at the edge, investigating if users? equipment is functioning as expected on the network, while improving QoS and QoE as networks transition to 5G. We would pull in one of our key partners that does analytics for RAN and for smart phones and can look at all the attributes of roaming from tower to tower, and communications between those two towers, all the different aspects of that phone connection. Drops, unnecessary packet losses, application performance and all those are things we can track at the edge vs. at the cloud. Having REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |33| real-time analytics at the edge can enable real- time load balancing, content delivery, application performance and other key items 5G is promising to the end user. We would bring in that partner, introduce it to the operator, and say, ?Our partner has already been validated. Let?s do the proof of concept. By the way, this is built on Wind River, Tieto, or Saguna?s software, which is also pre-validated and pre- integrated within the system.? We?re not just coming in with a piece of hardware. We?re coming in with a piece of hardware plus our ecosystem of partners that?s either embedded, referenced, or validated on the system. That de- risks the operator?s position. Monica: The operators also need help with monetization, because that?s a powerful driver to edge computing. How can you help them with this? Jeff: There are a lot of buzzwords that operators hear. You hear vBRAS, and vCPE, augmented reality, and now artificial intelligence, plus IoT. We try to listen to our customers. We want to understand what problem they?re trying to solve. If the problem is ?I want to make more money,? then we can bring in joint solutions with partners, or provide from our own experience examples of how operators can monetize their network. We try to work together as partners to understand what the hot buttons of their customers are. What are the key driving forces? What are the TCO and ROI they need? How can we help them, by integrating all the software within our hardware and becoming a one-throat-to-choke, where we?re the support entity or integrator? How can we help them if they go to different customers to test it all out? What?s also cool is that every time we go meet with a service provider, we?ll come up with different use cases. When we walk away, we?ll come out with 50 more use cases that we had never thought of. It?s really great. Then we go back to our offices and we look at them. For instance, ?We never thought about this. How do we integrate IoT in the transportation realm for a municipality?? Or a carrier might want to know, ?How do I operate my network within smart-city trials with ADLINK products and ADLINK?s ecosystem software products?? Because we?re at the forefront of mobile edge compute, we hear all these use cases of what we could do with them. Again, we bring that experience to other carriers in Europe, in Asia ? and all that gives us exponentially the different use cases. Monica: Your experience in different verticals helps, because a lot of MEC applications are going to be for verticals or IoT ? it?s more than just music and video caching. When we talk about MEC, we often talk about 5G too. What?s the relationship between the two? Is 5G going to enable some part of MEC that we cannot deploy now, or is MEC going to be necessary to support 5G? Jeff: I think it?s the latter. I think we?ll start seeing more and more MEC deployments, because MEC is a key use case for 5G. There are three to four different factors around MEC use cases for 5G. One is that we know the key deliverable of 5G is speed. How do I increase the amount of data throughput to my end user? Whether that end user is an appliance, an intelligent car, or a human, it?s all about speed. Instead of megabits, it?s gigabits. Second, you have the enormous number of devices that will be attaching to the mobile network. 5G has to be able to manage this. Where MEC comes into play is early on with the analytics, knowing how the network evolves as it is growing. Operators need to know how to use analytics to help them conform their network in certain areas, whether metropolitan, urban or rural. How can I use analytics to do that at the edge, at the tower itself? Using that data, I can say I need more compute power in this area to do this, this, and this for speed or because of the number of new appliances that are being attached. And later, the question becomes how to use things like video caching, content delivery. Areas that take up a lot of bandwidth, that require a lot of high processing speed for Xeon processors, that could be done in the cloud but need to be done at the edge to ensure that the user?s experience is there. MEC is an introduction into 5G. As 5G rolls out, I foresee more and more edge, cloud, and computing technologies to enable that speed. Once 5G standards are written, the radio equipment is pretty much set. Monica: What is ADLINK doing to prepare for 5G? REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |34| Jeff: We?re not selling radio equipment. We?re not selling some of the 5G components that are focused on the bandwidth and the frequencies. However, what we are selling are the high compute technologies at the edge. Again, that will be in the central office, at the customer premises or at the radio tower. A lot of those products are related to the virtualized state of the radio towers. As you start getting into 5G, you?re going to see both cloud RAN and virtual RAN. Our set-top box, which is our brand, is focused primarily on vRAN technology ? a cloud environment at the tower. ADLINK has also developed the Modular Industrial Cloud Architecture that enables a modular type of compute system that can be put into the central office, or a data center, or the customer premises. This architecture can support things like vCPE, deep packet inspection, security, load balancing, load routing ? all of these different attributes that need a lot of different compute power or I/O functions. We?re also heavily into open architecture. Our edge compute products are a modular cloud architecture. We?re submitting specs to the Open Compute Project, or OCP, specifically for telecom. It?s called Carrier Grade Open Rack. We truly believe in an open architecture that enables our partners or even our competitors to build the same type of components that are modular, that can fit together. Those are the key aspects of 5G, because once 5G starts happening, you?re going to see more openness, more suppliers like ADLINK and ecosystem partners needing to work together more, not only on the hardware, but the components for end-to-end connectivity, and the service training that goes with it. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |35| About ADLINK Technology ADLINK Technology is enabling IoT with innovative embedded computing solutions for edge devices, intelligent gateways and cloud services. ADLINK?s products are application-ready for industrial automation, communications, medical, defense, transportation, and infotainment industries. Our product range includes motherboards, blades, chassis, modules, and systems based on industry standard form factors, as well as an extensive line of test and measurement products, smart touch computers, displays and handhelds that support the global transition to always connected systems. Many products are Extreme Rugged, supporting extended operating temperature ranges, and MIL-STD levels of shock and vibration. ADLINK is a Premier Member of the Intel? Internet of Things Solutions Alliance and is active in several standards organizations, including OCP, ETSI MEC and NFV, CORD, OpenFog Consortium, TIP, the PCI Industrial Computer Manufacturers Group (PICMG), the PXI Systems Alliance (PXISA), and the Standardization Group for Embedded Technologies (SGET). For more information, please visit www.adlinktech.com. About Jeff Sharpe Located in Portland, Oregon, Jeff Sharpe is the Senior Product Manager for ADLINK?s Network and Communications Portfolio focusing on product investment strategies, Business & Market Development for the NFV/SDN/MEC/IoT markets. Jeff has over 34 years of Telecom and Data Communications experience with the key focus on network evolution strategies and product delivery. Prior to ADLINK, Jeff was Senior Manager of Solutions at Radisys and Managing Director of Next Generation Platform Products at Nortel Networks. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |36| Profile Advantech Since 1983, Advantech has developed hardware for a wide range of environments, verticals, requirements, IoT applications, and device types. Verticals include process control, digital health, power and energy, as well as smart city applications; the IoT applications are for industrial automation, intelligent systems and connectivity, among others uses. Because of the diverse markets it serves, Advantech offers multiple equipment types ? including systems, platforms and network appliances, all the way to embedded devices, which connect sensors, actuators and cameras, and rugged mobile devices (e.g., for transportation applications). MEC and edge computing fall within the scope of the Networks and Communications Group. Advantech provides hardware solutions at the outer edge that enable MEC and OpenFog deployments and the push to move processing and storage to the far edge. Advantech has recently introduced a virtualized platform, Packetarium XLc, for MEC and other edge-computing deployments. The platform provides a carrier-grade interim solution for 4G networks that is ready for 5G, and it is based on an open architecture and industry standards. Advantech positions it as a micro-datacenter-in-a- box, or a microserver with a modular design meant to provide scalability. According to the company, it can accommodate 9 slots, up to 288 Intel Xeon processor cores, and is typically installed away from big, centralized data centers. The carrier-grade Packetarium XLc supports five nines availability, complies with NEBS Level 3 and uses a 6U compact form factor with reduced depth (400 mm). The Packetarium XLc PAC-6009 flagship- model server is also designed to help operators transition from legacy solutions such as ATCA in a cost-effective way. The Packetarium platform is part of Advantech?s NFV Elasticity initiative that leverages scalable Intel-based platforms to enable service providers to deploy VNFs anywhere in the network ? and specifically in edge locations where proximity to the user improves performance, QoE, and resource utilization. NFV Elasticity helps operators integrate the infrastructure in the core and at the edge with RAN elements such as access points, macro cells and small cells. Advantech participates in an ETSI MEC PoC, ?Multi- Service MEC Platform for Advanced Service Delivery,? along with Brocade, GigaSpaces, Saguna and Vasona Networks. The PoC demonstrates how a unified NFV infrastructure with a cloud orchestration system can concurrently support multiple MEC platforms and applications. The MEC platform provides operators a way to leverage analytics at the edge, helping them to optimize RAN performance and, especially, to minimize latency. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |37| Advantech Scalability and flexibility at the edge A conversation with Paul Stevens, Marketing Director, Networks and Communications Group, Advantech Monica Paolini: To support processing and storage at the edge, operators need hardware that is scalable, cost effective and flexible. The requirements are variable, driven by a wide range of environments and performance targets. In this conversation, we talk about how hardware designed for the edge can meet these challenges with Paul Stevens, Marketing Director, Networks and Communications Group at Advantech. Paul, Advantech has a broad portfolio of solutions. What is Advantech?s focus on edge computing and MEC? Paul Stevens: Advantech is a manufacturer that?s been in the embedded computing business now for over 30 years. We?ve grown over that time to be a billion-dollar company, focusing on a diverse range of industrial areas. Over the years, our products have been used to connect thousands of industrial devices, sensors, actuators, and so on at the edge of the network, where we?ve been active in IoT-like applications such as SCADA and remote monitoring for a number of years. Various divisions across Advantech work on embedded computing in all of its forms. The group that I?m working for, the Networks and Communication Group, focused over the last 15 or 20 years on telecom infrastructure and enterprise security. We?re finding now that a lot of the businesses are beginning to converge as the new IP infrastructure comes into play. Equipment that we?d previously been designing for the core is now finding itself out at the edge as well. And there are different constraints there, as mobile edge computing, fog computing and so on start to come into play. Monica: What are the requirements that change as you move from the core to the edge? How do these changes affect you? Paul: I think there?s a need for more scalability and flexibility in platforms that we design for the edge. A couple of years ago, when NFV came along, we were somewhat concerned that everything would disappear into data centers and into the cloud, and we?d all be out of business. On the contrary, though, we have found that, continuing on our embedded tenets, there are a lot of opportunities and a growing market demand at the edge. Over Advantech?s history, we?ve acquired precious experience working with very stringent environmental specifications in central offices, for carrier-grade and NEBS compliance. Now we?re putting that experience to work in the harsher and more rugged environments of the systems we provide at the edge. On the scalability side, scaling out offers more processing headroom for example to add extra baseband and MEC application processing as needs evolve. Scaling up and down involves greater design flexibility and providing more cost- optimized networking gear to match precise workloads at specific physical locations. Monica: There are clearly different physical constraints. What about equipment size, or power? Does it depend on where you decide the edge actually is? Paul: Yes. It?s all about defining where the edge actually is, because it is a bit of a moving target. And it will continue to be a moving target over the next few years as technology advances. In mobile edge computing, for example, a lot of the discussions were initially focused on putting more intelligence at the radio head or the eNodeB. We?re beginning to see that bringing the compute or virtual edge up into a higher-level aggregation point is probably a better way to plan the network architecture and topology. The devices that we?ve been building are more ruggedized. If they?re for outside use, then there are temperature and size constraints, as well as power constraints to consider. If we?re moving into areas that are closer to central office requirements, then we?ve got to adapt to their various environmental needs and space needs. If we?re close to transmission equipment, we need short depth boxes and so on. The products that we?re designing scale from the edge of the network with a few cores, to these aggregation points, where higher levels of computing density are required. At the same time, we have to meet tight packaging and environmental specs. Monica: As you said, the edge can be a moving target. And you offer solutions for different edge REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |38| locations. But how do operators decide where the edge is for what they?re trying to accomplish? Paul: It will depend on the services they?re trying to offer. We can draw an analogy with customer premises equipment in the enterprise, where vCPE and technologies like SD-WAN are beginning to take over. There, the edge is moving between that customer premises equipment, and the managed services in the cloud. Now you can move those services where you want, from the cloud down to the CPE, and have them typically running on the CPE itself. We?re looking more at deploying devices that are closer to the user, that are more programmable, that are more flexible. We can now run virtual functions either on the devices that are at the edge of the network or up in the cloud, wherever it makes the most sense from a security perspective, from a performance perspective, and also from a latency perspective. Monica: The closer you go to the user, the lower the latency. But what?s the security advantage in moving functionality from the core to an edge aggregation point closer to the RAN, or to the RAN? Paul: The security gateway function in the RAN is evolving, especially with densification. We?re finding that it?s better adapted to aggregation points at the edge which secure data as it hits the network instead of placing gateways in the core. This can also take place in the same system as vRAN and MEC processing. Monica: What is the relationship between MEC and virtualization? Is a virtualized network a prerequisite for deploying MEC? Or do they complement each other? Paul: I think they go together because the whole ecosystem evolves around NFV. They are really playing hand in hand. From a MEC perspective, all the layers are built on multi-core processors where virtualization becomes absolutely key. Most companies are pursuing Network Functions Virtualization and service function chaining as a way forward. Monica: Within a virtualized network, you may have multiple edges, depending on the application. This gives the operator more flexibility in managing applications. Paul: Yes, the elasticity comes with being able to move those applications either from the data center to aggregation points, or even closer to the user entity. And just where is the edge when user entities and devices such as connected cars start to talk to each other, or make use of other neighboring devices for communication purposes? It becomes more difficult to decide just where the edge is. I think we?ll see it moving around, and elasticity is key to being able to move network functions around with it. Monica: What exactly is edge computing? What does it entail? MEC is one take on that. Then there are other initiatives such as OpenFog, or CORD. How do you see them playing together as we push the functionality toward the edge? Paul: There are a lot of open initiatives ? we?re trying to democratize, to a larger extent, what can REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |39| be done at the edge. Mobile edge computing offers a new ecosystem and a new value chain. The idea is for operators to open up their radio access network to allow rapid deployment of new services. We?re still in a period of transition. In 2017 we?re working on proofs of concept. We?ll start to see more trials out there. I think it?s key that we test out the various technologies to see how each of the new initiatives come together for a better solution. At Advantech, we work very closely with the vendors that are, let?s say, closer to the solution ? and that are working on the standards. To position Advantech, we?re not the solution provider ? we?re the building-block provider. Our expertise really is in the compute engines. It?s within the embedded products that we are capable of putting into these diverse locations. Monica: You mentioned scalability before. What?s your approach to providing scalability to your customers? Paul: We can scale equipment designs to meet performance, throughput, and connectivity needs wherever necessary. Typically, we optimize platform designs to meet those requirements and overall financial constraints. As an example, if a specific open and universal appliance priced at $300 is being deployed in 50,000 units, then, obviously, there are dramatic capex savings when compared with the deployment of an over-specified standard server equivalent that could multiply capex by 5 or 10 times. It?s very important that we work with our customers ? the OEMs, the system integrators, or the operators ? to try to tune those performance requirements for various locations in the network. Elasticity will allow us to move some of the compute capacity around. But in many cases we?re looking at what optimum platform performance is needed at specific places in the network. Monica: Virtualization and edge computing also change the ecosystem composition and relationships. It?s not just a hardware change. It?s also the way you run an operator network, and the value chain dynamics. How is edge computing changing Advantech?s relationship both with vendors and OEMs and with service providers? Paul: With NFV much has happened within the ecosystem. NFV has brought a lot of innovation. It does mean that there are many more building blocks to actually put together, hence the number of initiatives to standardize around NFV in various domains. For us, there is no blueprint right now. I think the ideal situation is to be, as we are, working very closely with select members of that NFV ecosystem, including virtualized network function vendors themselves. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |40| A lot depends on the operators. Some tier ones have their own R&D capabilities, others are doing the profiling and testing themselves. Then it will depend on the business model that best suits them for larger-scale rollout. In some cases, we?re either working with the operators? preferred system integrators or service vendors. In other cases, the operators are doing a certain amount of work themselves and we deal directly with them. The environment is quite flexible now. Some of those models are beginning to shake out as various NFV applications are deployed, particularly on the SD-WAN side. The important thing is to bring the ecosystem together and make sure we set up the trials, the testing of all the different elements, so we have solutions that are proven and ready to go. Monica: Do you think that with virtualization, operators will start taking more of a hands-on approach in selecting vendors? Does that mean you may be working more with service providers directly, rather than through partnerships? Paul: There?s a mix. Over the last year or so, several of our OEM customers have become our partners. They?re relying on us for hardware development as well as certification. In the past, they would take on certifications like NEBS for North America. Now, where there?s more of a partnership hand-off, we?re taking care of some of those tasks that they did originally. That?s from the OEM side. The operators are trying to take a new approach with NFV. We?re seeing more of a white-box approach, where we?re looking at a number of almost universal platforms at different performance levels. At the end of the day, it?s really how much performance you can pack into the space available, how much connectivity and offload you need in the system. We?re being influenced directly by the tier-one operators in that respect. Monica: You mentioned trying to get the ecosystem going. Proofs of concept are a great way to explore what is achievable. What is your experience to date on that? Paul: Pretty good so far. We kicked off a MEC ETSI PoC back in September. We were part of the Brocade initiative, along with Saguna and Vasona, as well as GigaSpaces with their Cloudify product. We demonstrated, an advanced service delivery proof of concept, at the MEC Congress in Munich. In the PoC, we put together a platform that makes it easier for developers from the application developer space to start developing applications they can easily and quickly be tried out in a live PoC or a live trial. The PoC is coming along. The results haven?t been published yet. We?re still in the evaluation and development stage with some of the APIs. I think over the next three or four months, we?ll start to see the results of that. Monica: What were the learning points in the PoC? What applications have you rolled out? Paul: The learning points were in bringing the various ecosystem parts together and working as a team to create a fully functional multi-vendor, multi-application MEC platform that allows both virtualized network functions and cloud services to be instantiated and delivered. We?re working on MEC user-plane functionality following the 3GPP documentation, as well as developing the various MEC APIs between platforms. Some of the partners have more experience in the orchestration portion of the NFV infrastructure. We?re bringing user plane functionality and APIs together so we have a platform that can accelerate the development of next-generation apps. From a MEC perspective, the new services and apps we?re looking at are connected cars, IoT, virtual reality and augmented reality applications. For that to start to happen we need to put platforms out there that are ready to go and can be connected up to a cellular network. This makes it possible to start smaller-scale trials sooner. Monica: It?s interesting that you mentioned IoT. At the very beginning, attention on MEC was mostly focused around video traffic optimization. Now there is a growing traction from enterprise and vertical-specific applications, as well as IoT. What that means is that the ecosystem also has to include partners from enterprise and, within that, specific verticals. The involvement of the enterprise is going to be crucial. Paul: Advantech has several divisions, already, that are working on multiple types of IoT. We are beginning to bring our groups together for a broader offering. There?s certainly a level of convergence that wasn?t there a year ago but that is now picking up REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |41| pace. Several of these platforms, and in particular the Packetarium XLc, can be seen as a micro data center at the edge of the network or extending the cloud. That?s where we?re going to find a new generation of data centers: at the edge of the network, handling in one particular aggregation point perhaps several thousand IoT gateways, connected cars, video surveillance cameras, and so on. Where there?s a need for fast response and latency, that?s where we?ll see that level of connectivity and local processing going on. One important factor though. While the failure of a single system in a data center hosting aisles of NFV infrastructure may not be critical, the situation changes as equipment gets deployed at the edge. In small sites with half a rack or less of NFV infrastructure, failure of a system can have a big impact on service availability and user experience. Edge sites are unmanned, remote locations meaning higher MTTR and service cost, compared to central offices and telecom data centers, so availability of each NFVI at the edge of the network is critical. Packetarium XLc supports high availability via redundancy at all levels via 2+2 redundant power supplies, the ability to withstand a single fan failure, redundant system management, system fabrics switches and control modules. Hot swap support for field replaceable units guarantees maximum uptime. Monica: What is unique about Advantech?s contribution to the MEC ecosystem? Paul: I think it?s providing the platforms that are adapted for MEC and needed now in the field. A lot of work in the various PoCs has taken place on standard IT servers, on data-center servers. We?re focusing on the edge and there?s a point where you need to go out there and start real testing. We?re listening to all of the different performance requirements of the players in this space. We?re working hand-in-hand with a growing ecosystem, to put together a platform that developers can get out and start working on. For example, at Mobile World Congress 2017 we?ll be hooking up some remote radio heads to the platform in a vRAN environment. Packetarium XLc, Advantech?s edge computing platform, brings all of that together. It can connect to the radio heads, perform baseband and MEC application processing, and also provide backhaul into the core network. We?re trying to put together platform configurations that are very scalable and very powerful so they can connect to many remote radio heads in dense areas. The platform can obviously also connect to small cells, as well, in a MEC scenario. At Mobile World Congress 2017, we will be demonstrating some of the software-defined radio applications that can be employed. Visitors will be able to discover how connect our micro data centers in a box to a real network. Monica: Today we are still at a proof of concept stage. When should we expect a commercial deployment of MEC applications? Paul: I think in 2017 the baseline infrastructure will come together for real testing and trials. Obviously, everybody?s got their sights set on 5G, which is still a few years out. In 2017 we?re also going to see more advances in new technology which can be deployed, and upon which more live applications can start to be tested and validated. Monica: What will you be doing at Advantech over the next five years? Paul: Our focus is going to be on adding services where required. Probably some higher levels of integration, depending on how fast the ecosystem changes move forward. We will continue to working closely with select ecosystem players. We will as such continue to bring together the unique expertise that we have in developing hardware that fits out there at the edge, and the diverse expertise of the other members of the ecosystem. At Advantech today, we position ourselves as the application-ready platform providers. As things continue to evolve, you?ll see us moving more up the integration chain to facilitate that process. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |42| About Advantech Advantech Co. Ltd. (TAIDEX:2395) is a leader in providing trusted, innovative products, services, and solutions. Its Networks & Communications Group provides the industry?s broadest range of communications infrastructure platforms, scaling from one to hundreds of Intel? cores, consolidating workloads onto a single platform architecture and code base. The group?s technology leadership stems from its x86 design expertise combined with high-performance switching, hardware acceleration and innovative offload techniques. For the new IP infrastructure, Advantech?s NFV Elasticity framework extends NFV to the mobile edge by supporting scalable carrier-grade platforms that run VNFs anywhere in the network. Operators, integrators and software- vendors can then rapidly validate the latest NFVI for vE-CPE, SD-WAN or MEC and benchmark VNFs using Advantech?s Remote-Evaluation-Service. For more information see www.advantech.com/nc About Paul Stevens Paul is Marketing Director for Advantech?s Networks & Communications Group. Paul has focused on technology marketing roles since he joined Advantech in 2002. Prior to that he was European Marketing Manager at Motorola Computer Group where he managed partner initiatives and helped evangelize new technology introductions. He is actively focused on helping build out Advantech?s NFV ecosystem network. He studied Electrical and Electronic Engineering in the UK and now lives with his family in France where he says the food is better. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |43| Profile Artesyn Artesyn has worked on developing power conversion and embedded computing technologies since 1971, initially as the Embedded Computing and Power unit of Emerson Network Power, and later through the acquisition of the Motorola Computer Group, Force Computers and Astec. Artesyn solutions span multiple verticals, including communications, broadcast and networking; industrial and transportation; and military, and government. In telecommunications, it provides hardware and platforms for the end-to-end network ? from the edge to the core ? supporting broadcast and media services and providing security and integration across technologies. Over the last few years, it has focused on virtualized solutions, both in the RAN and in the core. Currently it is developing the next generation of hardware to support the transition to 5G. The MaxCore platform is designed to meet service providers? needs for edge computing and network virtualization. With the MaxCore platform, Artesyn provides a fully integrated suite of cloud-based products that is scalable, flexible and power efficient. MaxCore virtualizes L1?L3 baseband processing in the RAN and supports MEC deployments. It is geared to optimizing performance and minimizing latency in environments with a high density of traffic and subscribers, thus targeting network operators? densification efforts and their transition to advanced LTE functionality and 5G. Within its commitment to fully virtualized solutions, Artesyn sees MEC as a solution it supports to expand the reach of networks, maximize performance, reduce costs, and generate monetization opportunities. Artesyn?s MEC server is designed to cover as many as 96 2x2 MIMO LTE-TDD 20 MHz sectors. It was developed in partnership with Intel and Wind River. Artesyn sees virtualized RAN and MEC as part of a flexible RAN approach that enables operators to reduce costs and to support a wider range of services for enterprise customers and dense indoor coverage. With the decentralized cloud data center approach that is central to MEC, operators can more efficiently provide bandwidth- intensive and real-time applications within the existing RAN infrastructure. Artesyn has worked on many use cases that require edge computing, including: ? IoT applications for smart cities ? Security gateways for residential multi-tenant uses ? OTT video delivery and transcoding ? Augmented reality ? Location-aware services for retail, government, health and education verticals ? Edge-based analytics to optimize RAN and application performance REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |44| Artesyn Edge applications and services to get closer to the subscriber A conversation with Linsey Miller, VP of Marketing, Artesyn Monica Paolini: Moving processing and storage to the edge is crucial to delivering the quality of experience that real-time applications require. It also enables the rollout of new revenue- generating applications and services. In this conversation, Linsey Miller, VP of marketing at Artesyn, shares her perspective on MEC, and tells us what she learned during the initial work on edge computing. Linsey, with edge computing, mobile operators and other service providers are looking at ways to improve performance and quality of experience for their subscribers by moving functionality to the edge. At Artesyn, you?ve been working on this area for quite a long time. Can you tell us what you have done to date in this area? Linsey Miller: Artesyn has been part of the wireless infrastructure for quite some time. We provide embedded computing and power conversion products to telecom equipment manufacturers and service providers, and our platform started to trend toward technologies that needed to be located at the edge and to support real-time functions. And so we got involved in a very dense compute platform we call MaxCore. MaxCore lends itself well to mobile edge computing, because it deploys multiple different virtual network functions and puts a lot of processing power into a small space. We?ve always been doing that in one way or another at Artesyn, from platforms down to blades, drawing on our Emerson network power and Motorola lineage. We?ve always been involved in serving the wireless infrastructure, so edge computing is an area we?re really excited about and honored to contribute to. Monica: You started working on edge computing before MEC even existed. You have your bases covered when it comes to mobile edge computing. Have things in edge computing changed since you first started working on it? Linsey: Yes, definitely. Before, a lot of functions could exist, but in a slower, more disconnected fashion, so you had to rely upon the core infrastructure for data plane processing, instead of provisioning that at the edge. You inherently had a network that was not really provisioned as efficiently as it could have been. This is one of the biggest merits of NFV, Network Functions Virtualization. Without NFV, you just couldn?t set up and tear down distinct services that you now can by having that capability at the edge ? that is, at the location where the user wants to use it. Monica: This allows operators to allocate their compute resources as they need them. Linsey: Right. This excites us because it means we could potentially save them a lot of money. If you?re provisioning for peak capacity, you?re putting a lot of equipment in a location that may not need it all the time. An example would be a shopping mall, a sporting event or a concert venue, which may be completely full of users one minute and may be completely empty the next. With mobile edge computing, we can now provision a network which moves to where users are going and which maybe even changes the functions within it based on the services they want to use while they?re there. Monica: In a dense environment, such as a stadium, subscribers want to do lot of video streaming and video uploads, and those are challenging from an infrastructure point of view. How can operators deal with that? Linsey: We?re focused on a couple of things: density and latency. We are the hardware provider, and with mobile edge computing, the promise is that you can do so much more at the edge with all these unique and innovative applications, but on general purpose hardware. We?re taking that hardware and making it serve more users, with as many services as possible in one place depending upon where they go. Within our platform, we are focused on reducing latency: the communications and the compute functions that are within that platform enabling all those applications are happening as quickly as possible. That could be the difference between seeing a live video streamed from a drone, versus sitting there and waiting for your Instagram picture or video to upload or download from the network. The other focus at Artesyn is on density. We?re trying to get as many users and as many cells as possible into a small space working off one edge platform. Monica: Stadiums are a good example of a venue where you have both density and latency REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |45| requirements at the same time. Can you tell us about trials you have done in this area? Linsey: This started for us officially in 2015, and then at Mobile World Congress in 2016. We were excited to partner with China Mobile, which was one of the first operators to be as outspoken as possible about mobile edge computing and virtual RAN, along with Wind River and Intel. That year we also did a demonstration with Telefonica and SK Telecom. Since then we?ve been involved in various demos. One was with China Unicom and Baicells, which did a drone demo. We?ve also been involved with Deutsche Telekom at the NGMN conference. NGMN has been the source of a lot of great info on mobile edge computing and 5G. We?re excited to be a part of a lot of different initiatives with operators, because many of them are being innovative about this and really getting hands-on about how they can take advantage of this, because it means they can offer new services. Another example is with Verizon. Verizon recently did an Innovation Lab challenge, where they invited multiple technology providers to come in and show some new services. One of the categories for that competition was low latency, and we won a low latency award with our MaxCore platform because we were, essentially, showing a mobile edge computing application. We were showing 360-degree virtual reality in the context of a sporting event. Imagine you?re at a stadium, you?re watching a game live, something just happened that you want to see a replay of ? you could actually hold your phone up and see a replay of that, with a 360-degree view. This a great example of new services operators can deploy that users will just come to love, and eventually not know how to live without. Monica: This is a very good example of a new service. What other new services are operators looking into that are more than just streaming video up or down? Linsey: When you look at some of the virtual functions that mobile edge computing can put in place, it?s exciting from a consumer point of view as well as an enterprise point of view. We worked with one of our application partners, Clavister, to show virtual security gateways. We can show over 3,700 virtual security gateways being enabled by a single platform. At the mobile edge computing conference and we saw Carnegie Mellon show one of its examples: a virtual network function based on facial recognition. That?s really exciting when you think about how a location-aware element can identify users and know them and cater things to their needs. Augmented reality, ad services and ad monetization are also great ways operators can use mobile edge computing to enhance the user experience. The drone demo is from another trial we did. You can imagine how video footage from a drone at live events is a great application that can be REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |46| powered by the mobile edge computing network, as well. The focus today is on which of these services can help operators make money now. Those are the ones where customers and operators have the keenest interest in how can we provision the service. But there are a lot of new, exciting services for the future and for 5G. Monica: Local services can support monetization efforts ? for instance, advertisements or applications that deliver content tied to a specific location. Linsey: Yes, even concessions ? it could be something that simple. But if you think about Pok?mon Go as an example of augmented reality and how you can enhance a user experience while they?re shopping, maybe seeing a game or something like that, you could make it really fun and you could make it something users really want to buy into and really want to use while they?re there. Monica: Are MEC and NFV changing Artesyn?s position within the value chain? Linsey: They are changing our position in the ecosystem a little bit. They also give us another opportunity to show how our platform can really help drive MEC and NFV forward. When you think about NFV, you think about dynamic provisioning and how you could do that better as a general application. But you can also think about the exciting possibility of enabling multiple different virtual network functions on one platform. Before, you had multiple different bespoke appliances, with each doing a different thing; now you can bring those into blades and a platform, and they can be doing different things, and you can turn them on and you can turn them off. Network slicing is the amalgamation of all those different virtual network functions. We think it makes easier for operators to deploy new services, because now they can go and say, ?I want to pick my very best application solution providers. And then I want to have this general-purpose hardware that I know is going to be high performance, I know it?s going to have low latency, I know it?s going to have high density. I can support lots of different service instances, and I can do them quickly. I know if there?s a real-time application I want to incorporate, this one is going to work for me.? We?re trying to build that rock-solid platform that can go in all of those different directions, and then work with the various different application solution providers to benchmark and show how it?s real. As a company, traditionally we have sold to the telecom equipment manufacturers ? the appliance vendors ? but because of NFV, our role is changing. We go into a position where sometimes we sell directly to the service provider, but we are also working together with that telecom equipment manufacturer to make sure its application software is running on our platform in the best possible way. We?re open to both models, because ultimately those telecom equipment manufacturers could be the super integrators of all those different things, and in some cases they may REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |47| just empower the software that runs on our platform. However, what makes Artesyn unique in the market is that we are an integrator for complex workloads. Whether it?s a real-time function, or a mix of different types of resources, or different types of compute resources ? they could be Intel, could be NVIDIA, could be FPGAs, could be camera inputs ? we bring all of those things together in one platform that can go in all those different functions. Monica: You may end up working more closely with operators, because they have started to take a more active role in building and operating. Linsey: We?re excited to work with some of the ones that are really innovative, because they know which applications are going to help them monetize those services. They have very specific problems to solve, and our team acts as a good extension of their resources when it comes to that. We?ve been really thrilled to collaborate with so many of them already on mobile edge computing. Monica: Is NFV a foundation and a requirement for MEC? Linsey: I think they go hand in hand. Some of these problem areas are great candidates for mobile edge computing, the way they?re deployed today. For instance, in an enterprise or stadium, there may be a DAS and that is not their problem right now. It?s owned by the building owner or the stadium provider, and the operator can?t offer great services on its network because the DAS just doesn?t really scale. Mobile edge computing really does need NFV. You need to have equipment that is dynamic and can expand and contract based on what the network needs. I think there?s going to be a big sea change on ownership of the network as a result of mobile edge computing. If the stadium owner can take that overhead off its plate, and not have to buy and maintain that equipment, but can instead buy it as a service from the operator, and the operator can ensure that the network will serve up those ads and the content the stadium owner wants to stream to its consumers, then it?s the best of both worlds for both sides of that equation. The operator can now monetize services that it couldn?t before. And the stadium owner no longer has to maintain something, which isn?t its core competency anyway. Today it?s frustrating to the people that come to their venue, quite frankly, if their wireless experience is below their expectations. Monica: Stadium owners may also provide the same infrastructure for different operators, so there may be some consolidation at the edge. Linsey: Yeah, that?s really exciting, having an operator-agnostic framework there. Monica: That also might be very important for the business case, because with MEC you?re adding infrastructure to the network. So how can you justify the business case? Is the improvement in performance sufficient to justify the cost that MEC adds in terms of equipment? Will venue owners have to contribute to the initial cost? Linsey: I would think so, because there are areas where they can benefit from MEC being in place. Imagine if you?re at a sporting event and the venue owner can serve up not just concession ads, but also jerseys or things you would buy in a store that you otherwise might not physically see as you?re walking out, but it comes up on your app, and you can have it shipped to your house. All sorts of interesting follow-on purchases, ad monetization, or loyalty could happen as a result of that experience. Monica: Do you need to have MEC for 5G? Or do you have to wait for 5G before you realize the full MEC capabilities? Linsey: I would say, ?Certainly not!? You don?t have to wait for 5G to take advantage of MEC ? it?s here now. A couple of examples: We demonstrated 96 cells on our current platform, and that was LTE with Intel. We also partnered with a company called Amarisoft, and we?re getting up to 120 cells, and that?s 2x2 FDD LTE. With LTE, operators are now in this lull because they expanded their network, they made it better ? and equipment providers had a nice boon from that build-out ? and then things leveled off. LTE networks have so much capacity still, and can support many of these services that may have video. There are many new services that can be deployed on an LTE network with no rip and replace. I think there?s a keen interest from operators to take advantage of the existing infrastructure and build on it. MEC is a wonderful bridge between the LTE network that is there today and the 5G network that will come in the future. But we don?t have to wait for 5G to deploy MEC. Monica: How will 5G make MEC better and more powerful? REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |48| Linsey: It going to get really exciting in the context of what the users can do with 5G. With driverless cars, and machines that can communicate with each other, 5G is going to bring a completely different level of performance just based on how much more bandwidth it?s going to allocate for applications like that. There will be more real-time applications, and more mission-critical applications. With LTE we?re not going to have as many services that are life or death. It might be really exciting, cool and fun from a consumer point of view, but handing over a lot of that control into functions that you never before dreamed of being done without human interaction, that?s where 5G is really going to be a different experience for all of us. Monica: What are you working on at Artesyn now? Linsey: Our first stage for mobile edge computing was creating a platform and a technology that would enable you to do all of these different virtual network functions and virtual RAN, and the network would incorporate just the basic tenets of mobile edge computing. Now we?re working on taking that to the next level, to the mission-critical functions. We?re putting more features into our platform to enable very high reliability. We are looking at things that the telecom network has done historically at its core, and implementing those at the edge so you have high availability, high reliability. We?re also working on the incorporation of more and more processing elements that exist as types of resources within those platforms. So, things like GPU functionality, very dense computing, higher connectivity, and proving out all of those different applications with our different ecosystem partners. It?s a lot of work, but it?s exciting when you look at the operators that are moving fast in this space: they?ve got the use cases that will hit the market first ? that?s out priority. But in the background, we are an infrastructure provider and we integrate complex workloads. We want to make sure that all of these things are working together, whether they?re doing graphics processing, compute, or connectivity. That is our core focus, and just making that better and better, and more and more reliable at the edge is our design focus in the next couple of years ahead. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |49| About Artesyn Artesyn Embedded Technologies is a trusted leader in the design and manufacture of reliable embedded computing solutions. For more than 40 years, Artesyn has enabled customers in a wide range of industries including communications, military, aerospace and industrial automation to accelerate time-to-market, reduce risk and focus on the deployment of new features that build market share. Building on the acquired heritage of industry leaders such as Motorola Computer Group and Force Computers, Artesyn is a recognized leading provider of advanced network computing solutions ranging from application-ready platforms, single board computers, enclosures, blades and modules to enabling software and professional services. About Linsey Miller Linsey Miller is Vice President of Marketing for embedded computing solutions at Artesyn Embedded Technologies, leading a team that includes global product marketing, technical marketing and marketing communications. Linsey drives solutions and partnerships that enable Artesyn?s customers to succeed as their business models change by improving the performance, efficiency and capability of their critical computing infrastructure. She has previously held senior sales and marketing positions with Emerson Network Power, Interphase and Verizon. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |50| Profile Intel Intel solutions are geared to addressing three business and networking challenges that service providers face ? their need to ? Increase network capacity to meet growth in data traffic ? Build agile networks to accelerate innovation ? Protect data traversing their networks This requires an end-to-end transformation toward virtualized, software-defined and cloud- ready networks ? from the devices all the way to the core and the cloud. Driving this transformation are Intel?s silicon and software solutions, but Intel is also working to ? Advance open source initiatives and standards ? Build an open ecosystem with initiatives like the Intel Network Builders program ? Collaborate with service providers, cloud players and enterprises In wireless, Intel?s focus is on the virtual network infrastructure of 5G in three areas: ? Radio access technology, with anchor booster beamforming, NR technology, and massive MIMO solutions ? Access network, with FlexRAN, including solutions for C-RAN, vRAN, small cells and MEC ? Core network, with router, vEPC, backbone and network slicing solutions MEC plays a central role in Intel?s strategy toward virtualization within a unified management plane for service management and orchestration. Intel sees MEC as a technology that will decrease RAN?s opex, and that will help service providers create new services and generate new revenues. The company foresees a wide range of business opportunities in verticals such as healthcare, energy, manufacturing, retail, government, transportation, financial services, and education. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |51| Intel MEC gets ready for the enterprise and IoT A conversation with Caroline Chan, Vice President Data Center Group and General Manager 5G Infrastructure Division, Intel Monica Paolini: Virtualization gives network operators ? both fixed and mobile ones ? flexibility on where to locate processing and storage within their networks. And they have started to move beyond the centralized cloud and place some function at the edge, or near the edge. Multiple- access edge computing, or MEC, is one initiative that provides a standards-based framework to operators and vendors to enable edge computing. I talked to Caroline Chan, the Vice President Data Center Group and General Manager 5G Infrastructure Division, at Intel, about the impact of the move towards the edge of the network. Caroline Chan: This is one of my favorite topics, so I am excited to have this conversation. And I noticed that you used the right word: you said ?multiple access.? We used to call it ?mobile edge computing? when we started it, but now we?ve expanded the horizon. Monica: It is still somewhat difficult to get used to the new name, but it?s more than just a change in what the MEC acronym stands for. But before we get into the relevance of the move to ?multiple access,? can you give us an overview of what Intel is doing in MEC? Caroline: At Intel, I am part of a group that provides the silicon platforms for software to a whole host of companies in the communications industry. I am focused on wireless, specifically on 5G infrastructure. We started to work on MEC three or four years ago, when it was just a twinkle in our eye. We started this work with Huawei, Nokia, Vodafone, DOCOMO, and IBM. We kicked off this initiative in MEC in an attempt to cloudify the network ? as a way to bring the best-known methods to a cloud service provider, within the innovation cycle. This gives operators the ability to generate additional revenue and start capturing some significant savings. Now this is clear as the industry is moving towards more and more network function virtualization. But, initially, with MEC we had to ask ourselves, ?What are the benefits we?re really looking to have? What benefits can we reap from it?? Today, I am very excited. We started MEC and it?s evolving: we?re now talking about trials and early deployments. We?ve really come a long way, and we?ve been in this as a technology partner to our customers and end users from the very beginning. Monica: As you mentioned in the beginning, MEC stands for ?multiple access,? no longer ?mobile access.? It started off as part of the virtualization of mobile networks, and now the scope is expanding. Caroline: When we started, because our background was in wireless and everybody I mentioned is involved in wireless, it was natural to concentrate on wireless networks. But then within the ETSI MEC, the operators asked, ?Why limit this to cellular? Why not Wi-Fi?? So we said, ?Okay, let?s add Wi-Fi then.? But then people asked, ?Why just wireless? What about wireline?? And this brought about the second release that?s coming up. We redefined MEC as ?multiple-access edge computing.? Because once you realize what the benefits of MEC are, you see they are applicable to all types of access networks. Monica: It?s an advantage to have the same approach available across different networks, right? Caroline: Yeah, exactly. Monica: Let?s look at the future also. So today, we have 4G networks, and we started deploying MEC in 4G. But what about 5G? Are there parts of MEC that need 5G? Caroline: This is the conversation we?ve had from the beginning: people would say ?why?? and ?can we?? so we started with 4G, and now we are continuing with 4.5G. Most of our trials today sit on 4G and 4.5G. So what does 5G really bring you? A lot of applications we talk about require extremely low latency. And that?s where MEC really comes in. All the speakers at all the conferences I?ve gone to, and all the operators I?ve talked to, emphasize NFV as the foundation of 5G. With NFV, the investment that?s required drastically decreases, because the foundation is a virtualized platform that uses general purpose hardware. And the multiple- access benefits, the MEC benefits, fit into this. In 4G, MEC is something that is nice to have: it opens up your revenues, it opens up your ability to deliver services. But for 5G, MEC almost becomes a de facto requirement. We?re very excited about it, and the road to MEC already started three, four REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |52| years ago. We?ve really started seeing our deployment this year in 2017. Monica: This process is going to require time, because we have to learn many things to deploy MEC, and this requires a different mind-set. You mentioned the benefits of MEC. Can you go over what is the value of MEC, especially to the enterprise? Caroline: When we started this, we started with caching. Moving the content closer to the user makes sense to a lot of people. And, as we dig in more and more, that is still a valid benefit. Now, one of the things we?re realizing is that, once we go talk to the operator, a lot of times the enterprise use cases become very prominent. Internally, I call MEC the perfect marriage between IT and telecommunications. When you go to the enterprise, when you talk about launching applications, many times you are talking to the IT side of the enterprise. And IT has a different methodology, different principles and different personal methods than the telecommunications side. The IT team is much more comfortable when you hide that communications part from IT ? making it seamless. With MEC, we are saying, ?Here is an API, as defined by ETSI, and you don?t have to worry about all the plumbing underneath. With APIs, you have a rapid innovation cycle so you can start putting in your applications. And by the way, these are applications that IT can control. The applications that IT wants deployed. All using a virtual machine with security.? Regarding security, as part of our learning, we have gone through a lot of security issues, related to billing, lawful intercept. A lot of the things we worked through today became a play for the enterprise. Really opening up and transforming the business from a consumer-driven subscription model to an application-services enterprise model. Monica: This changes the relationship between the operators and the enterprise, and gives the operators a way to address the needs of the enterprise in new ways. Traditionally this has been an issue for mobile operators. Caroline: And, in fact, we see that some of our partners go to market through the channel model. If they package the MEC right, they can sell it from their channel through MVNOs. When we started our work on MEC, this was part of our vision; it was on our wish list. And we started seeing it coming true as we worked through all of the learnings. It does change the model for a lot of operators. It makes the operator more comfortable with the enterprise side of MEC, and capable of solving some of the B2B issues that ETSI is working on. Monica: Wi-Fi plays a crucial role today in the enterprise wireless network. How is MEC going to fit in? You can argue that Wi-Fi is in competition with MEC. Alternatively, you can argue that MEC gives the enterprise a way to share the same services across different networks, cellular and Wi- Fi. Caroline: We see Wi-Fi and cellular as complementary. We never view Wi-Fi as a competitor. Cellular and Wi-Fi can coexist. In the past, an enterprise would choose between Wi-Fi REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |53| and cellular. But we?re seeing enterprises go into MEC deployments using both. MEC has become access agnostic, and we are advocating a multiple access platform. You can use it either in cellular or Wi-Fi networks, as you see fit. Or even in wireline networks. MEC becomes an access-agnostic way of deploying applications. It?s right on the network?s edge, as close to users as possible. Based on what we have done, and our partners have done, MEC is predominantly deployed in LTE networks. But going forward, we absolutely see Wi-Fi being part of it. That?s why ETSI is moving rapidly forward to include Wi-Fi and wireline in MEC. Starting in late 2015, we have seen some sporadic deployment ? early deployment of trials ? and they have been predominantly LTE based. Monica: This is a question that you probably get all the time: What is the business model for MEC? Because from both an enterprise and an operator point of view, MEC requires an additional investment in infrastructure. How can we justify the additional cost? Caroline: That?s something we?re going embark on this year. Now that we have deployments out there, we want to get a TCO model. From the early indications we have, the payback time is promising. You roll out MEC and then you start to leverage your new ability to deploy multiple applications ? not one or two, but tens of different applications running on the same platform. This is a step forward compared to the past, when you might have had to roll out one platform for each application. The payoff looks quite promising, and we are working more on it this year. We would like to publish a MEC business case for our end users. Since we started working on MEC, we?ve been doing a lot of edge caching. That seems to be pretty straightforward in terms of being able to save on backhaul. But we didn?t stop there; we started looking at different verticals related to IoT. We asked ourselves about retail, transportation, health care and industrial. Each of them does have its own matrix of things to do, so we started attacking them one at a time. One of the well-known trials was with China Mobile and Nokia, on Mobile Formula One, and it was within the sports vertical. This is a rather straightforward use case, because you are simply giving your subscribers a way to see a sports event much more close-up and personal. At the same time, the operator gets a share of the revenue for video broadcasting. The TCO looks very promising. Different use cases have different matrixes and different returns. We are seeing very promising results in some of the verticals, and we are working with our partners and customers to start publishing this kind of information. Monica: As you mention, there is considerable variability across verticals and environments. Could there also be different sources of funding? For instance, since the enterprise or the venue owner stands to benefit so much from MEC, could they become willing to pay for the MEC infrastructure? Caroline: I think that?s very feasible. Especially if you are looking at all the new spectrum that is becoming available. Here in the US, there?s discussion around CBRS and lightly licensed REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |54| spectrum. It does create a different environment. Opportunities created thanks to new spectrum mean that vendors can sell new services and applications through vertical channels. Or what about an MVNO coming in and funding part of this? It doesn?t all have to be funded by the operators themselves. The enterprise may purchase some of the equipment under a sharing model. I think MEC becomes an enabler, and this makes the different ownership models interesting. Spectrum owners need to see a return for their capex. If MEC is proven to fit well within their business needs, it gives them an incentive to take that next step in becoming a participant in the capex. Monica: We?ve been mostly talking so far about MEC, but there are other initiatives, such as OpenFog and CORD. How do you see MEC within the context of these other initiatives? Caroline: Every time people say there?s a new initiative coming, it means there?s a heightened interest within the ecosystem to make edge computing happen. I think OpenFog, Open Compute, CORD, and MEC will eventually all merge, because they serve different segments. I think that?s what ETSI is trying to do. At the last ETSI meeting on MEC, we hosted an OpenFog day, where the two camps were sitting together and we learned from each other. Will they all become one initiative? I don?t know yet. We have seen multiple initiatives coexist addressing different parts of the market. But I do see some merging, especially as IoT becomes 50% of the market. That needs to be addressed from an enterprise perspective. But I think for now, all of the innovation that is coming from all camps only helps the industry to move forward. And to move faster. Monica: Could there be a risk of fragmentation or the development of proprietary solutions, in this respect? Caroline: I hope not. I guess there?s always a risk, but I think the camps are absolutely in dialogue. You know, in that day at ETSI, I learned so much from the OpenFog people, and there was all this exchange of ideas. So I don?t see a high risk of fragmentation. I see the camps learning from each other, maybe initially addressing different parts of the market. Instead, I do see a probable convergence at some point. Monica: Recently there has been a lot of interest in network slicing. And with network slicing, you manage traffic based on applications and services. This is something that MEC does, as well, because depending on the function, you do the processing in a different location in the network. This is different from what we have in the networks today, where all the traffic goes through the same channel. So what is the relationship between network slicing and MEC, or edge computing solutions? Caroline: MEC and network slicing complement and enable each other. In fact, the Intel Mobile World Congress 2017 demo will include network slicing and MEC. I don?t see them as competing with each other at all. Network slicing is an enabler for 5G, and is built on top of a solid and flexible platform. There will be different slices targeting the massive and varying IoT ecosystem, and enhanced mobile broadband. MEC can sit at each one of these slices, and we?re going to demonstrate that the two enhance each other very well. Monica: Intel has been quite busy with edge computing trials. What were the major lessons you learned? Caroline: Number one is, when we started, there were a lot of questions about the killer app. For a while, I couldn?t get my head around ?What is a killer app?? One thing we learned is that you securing the network is a fundamental priority, both in the hardware and in the software. I remember, we saw this for the first time when lawful intercept was introduced. Also, once you start doing the trials, once you start talking to different segments, different enterprises directly, you see a set of killer apps. There?s not one app that fits all. Every vertical has its own different needs. The most fundamental thing is to provide a platform that is secure, that is virtualized, and that has a fast deployment cycle. This is very, very important. That?s one of the hallmarks of a virtualized platform. And it must not impede the operations side. So, in other words, all your billing, all your OAMs, or operations, administration and maintenance, need to be packaged well. And then the other learning point is that you do need to take care of the channel. You need to make your packaging such that you can easily sell through the channels. You don?t need to worry about the killer apps, because once you go out there and engage, you will get the applications REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |55| that run efficiently, cost-effectively and beneficially on the platform. So now what?s my biggest lesson learned? For a while, we were brainstorming, we were doing surveys. And then we realized that before we found the killer app, we had to get a platform that people are willing to deploy. If you pass the security audit, then the application discussion can start. Monica: How do you see Intel?s role in promoting the growth of the ecosystem? Caroline: Intel has always played the technology- enabler partner role. For this, we have the Network Builder program. It?s part of our NFV initiative. It?s a way for us to get our ecosystem together. It includes a lot of types of companies, including software companies, component providers, system integrators and middleware companies ? and they rally us around MEC?s community. With the Network Builder program, we?re matchmaking. There is even some bucket-enabling funding. We also have a very large enterprise sales force. And we organize, and we work together with our partners. Ultimately, we don?t sell the system, we sell the technology. So we will innovate, and invest in growing the ecosystem. We?ll also do things like study the provider business case, the provider business model, and we?ll participate in the ETSI standard. We?re part of the OpenFog initiative, as well, and we help make sure that the standard is open and it adheres to the NFV principles so it really becomes an enabler of the ecosystem. Monica: What are you working on right now, and what should we expect from Intel over the next five years? Caroline: We?re making it well known that we are invested heavily in 5G. We tied MEC into our 5G initiative, and that?s why it?s part of my team?s mandate. We?ll work closely with partners and customers on it. We?re developing many solutions, including network slicing and a reference platform. We are developing a platform where we are advancing the ecosystem using network slicing. We are also making the MEC SDK software development kit available to the ecosystem. Anybody who wants it may come to intel.com. If you qualify, we will send you one of these. And we?re going be supporting and participating in more trials and proofs of concept. And really see this taking off in the industry. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |56| About Intel Intel (NASDAQ: INTC) expands the boundaries of technology to make the most amazing experiences possible. As the leader in the PC industry, Intel is powering the majority of the world?s data centers, connecting hundreds of millions of mobile and Internet of Things (IoT) devices, and helping to secure and protect enterprise and government IT systems. Our manufacturing advantage?fueled by our pursuit of Moore?s Law?lets us continuously push the limits of performance and functionality and expand what experiences can be made possible. Intel has a growing portfolio of products and technologies that deliver solutions to help communication service providers transform their networks, bringing advanced performance and intelligence from the core of the data center to the network edge. Intel?s commitment to network transformation is long and deep ? with years invested in delivering reference architectures, growing a strong ecosystem, and partnering with end-users. We are also deeply committed to 5G which represents the true convergence of computing and communications. 5G is a fundamental shift for the industry where networks will transform to become faster, smarter, and more efficient to realize the potential for the IoT and mobility, enabling richer experiences throughout daily life ? augmented reality, smart cities, telemedicine, and more. Information about Intel and the work of its more than 100,000 employees can be found at newsroom.intel.com and intel.com. About Caroline Chan Caroline Chan is the Vice President Data Center Group and General Manager 5G Infrastructure Division within Intel?s Network Platform Group (NPG). She is responsible for leading a cross functional organization driving global network infrastructure strategy for 5G. Bringing Intel processor into the wireless infrastructure, projects such as virtualized RAN, mini-cloud RAN, 5G network, heterogeneous network consisted of small cells and Wi-Fi, and mobile edge computing for IoT. In her role, she closely works with telecommunication vendors, operators, and application developers. Caroline also represents Intel at industry forums. Her research interests include 5G and HetNet performance. Prior to joining Intel, Caroline was Director of Product Management at Nortel Networks where she managed a portfolio of 3G and 4G wireless infrastructure products. Caroline was born in Nanjing, China, received her BS EE from University of Texas at Austin, and MS EE from University of Massachusetts at Amherst. Outside of her family and work, Caroline is passionate about the Texas Longhorn Football team. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |57| Profile InterDigital For over four decades, InterDigital has developed new mobile technologies and contributed to global wireless standards. The company?s activities are organized around the concept of the emerging Living Network, in which intelligent networks self- optimize to deliver services that are tailored to the content, context and connectivity of the user, device or need. In the edge computing space, InterDigital strives to bring MEC to the far edge ? small cells, HeNB and STB, as well as user devices such as personal devices or devices embedded in vehicles. Processing and storage at the far edge enable ultra-low latency, location-specific and real-time traffic optimization, agility and adaptability, as well as context awareness. These performance enhancements are designed to support distributed MEC use cases such as multimedia content, gaming, and personalized cognitive assistance, and generally to aid the development of applications that benefit from edge architectures. InterDigital actively participates in the ETSI MEC, as well as in the Open Fog Consortium and other open-edge initiatives, as well as in groups with a wider mandate, such as 3GPP, European Commission?s Horizon 2020 and NSF?s PAWR. It also participates in an ETSI MEC PoC in collaboration with Intracom, CVTC Limited and University of Essex. The PoC uses FLIPS, a multi- cast video technology designed to lower the latency in the transmission of real-time content. InterDigital?s MEC vision is rooted in two technologies that the company believes complement and enable MEC deployments: ? Flexible routing, which combines SDN and ICN and provides the foundation for FLIPS. ICN shifts networking away from host-to-host communications, to content- and name-based addressing. InterDigital believes that this paradigm shift is necessary to meet 5G?s latency requirements, by facilitating pushing content to the edge. ? Dynamic surrogates, which support ICN-based orchestration that allows operators to allocate network processing resources where and when they are needed, using softwarized servers located at the edge. The goal of surrogated services is to predict and dynamically meet traffic demand, taking advantage of the flexibility afforded by NFV. InterDigital continues to pursue its R&D activity to integrate services with the network infrastructure, as SDN, NFV and MEC converge in the evolution of 4G and the development of 5G. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |58| InterDigital Enabling low-latency, high-reliability apps with edge computing A conversation with Debashish Purkayastha, Member of Technical Staff, InterDigital Monica Paolini: MEC and, more generally, edge computing are crucial to ensuring that low-latency and high-reliability applications can be successfully deployed in 4G and 5G networks. Debashish Purkayastha, Member of Technical Staff at InterDigital, shares with us his thoughts on how this transition to mobile edge computing, through initiatives such as MEC and OpenFog, will change the way we operate and benefit from wireless networks, and how we will be able to meet the new latency and reliability requirements. Debashish, can you tell us what you and InterDigital are doing in mobile edge computing? Debashish Purkayastha: The development of the mobile edge network has the potential to lead us to the next killer app. There are mainly two reasons. The first is that edge computing can dramatically increase the responsiveness of the network. The second is that it gives application programmers access to mobile network information, which many of the developers don?t have at this time. We started to work on edge computing a few years back, when we identified the need for enabling the edge. This means adding capabilities at the edge and defining a range of services that can be offered from the edge, taking advantage of the network information as well as user information. We started by working on edge caching, storing content at the edge of the network, and then enabling video-distribution services from the edge with a more detailed knowledge about what the user likes, what the user?s preferences are, because those can be easily and accurately measured from the edge of the network. While working on that, we realized that we did not need to constrain ourselves to content, and that we could move computation, as well as storage, to the edge of the network. The first challenge that we faced was defining where and what exactly is the edge. Depending on the use cases, the edge may be at different levels of the network. Another problem is the availability of so many devices and platforms, which create a heterogeneous environment: devices of different shapes, different sizes, different capabilities. These devices may be owned by different parties and different network operators, and use different access mechanisms. It?s difficult to develop applications on such a wide range of heterogeneous devices. The third challenge is for the developers. With so many different devices, different access mechanisms, different owners, how can a developer write an application that runs uniformly across all such platforms? Monica: With edge computing, we gain the opportunity to manage traffic in a more intelligent and sophisticated way. But at the same time, we have to deal with a higher level of complexity. From an operator point of view, the increase in complexity can be challenging. What are the advantages? Debashish: The main advantage is the reduction in latency. With lower latency, we will see instant improvements in the responsiveness of applications, which in turn will improve the satisfaction of the customers that the network operators are serving. Another important aspect of edge computing is the reduction in traffic in the backhaul network. As small cells are being deployed, connecting a small cell to the backhaul network can be challenging. It is difficult to run fiber to each small cell, thus creating a bottleneck in the operator?s network. To avoid that potential bottleneck, edge computing provides a reduction in traffic ? because we have the capability to process users? requests and data at the edge. It will be possible to respond to user requests from the edge and avoid data traveling over the backhaul network. The third important aspect is privacy and security. As personal data traverses the internet, it is vulnerable to being stolen. With edge computing, we can process data at the edge, so we can keep the data in a local context. We don?t let it go outside the local context, into the internet. Many REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |59| users and applications demand that kind of privacy and security. Monica: You mentioned you may want to keep the content local to enhance security or to choose different levels of performance for applications. But who is going to decide? Obviously, the first candidate is the mobile operator, which decides how to manage traffic based on its network resources. But it could also be the enterprise that decides what content stays local. Or it could be the user or the content provider that decides. How do you see that working out? Debashish: By deploying MEC, we have more options for traffic management, or traffic control. Where does my data go? How far does it go? These are decisions that are available to the user, the network operator, the application developers, and the enterprise itself. Right now, we think it is a decision that can be application specific or user- requirement specific. There will not be a single rule, and the decision may vary depending on the application, context, etc. For example, in certain applications, such as a video distribution service, the user may be OK with certain kinds of low-quality video, and this will allow the network operator to lower the video quality to improve the efficiency of the network. In such cases, the operator may decide what data rate a particular user should be allowed, based on the utilization of network resources. For example, in the video distribution service, it may not be possible to give HD video to every user who says, ?I always want HD-quality video.? In that kind of application, the control will likely stay with the network operator. It is different, however, if we?re talking about user data, such as medical records that are being processed at the edge in a medical application. In that case, it may be the user who says, ?My data should be processed at the edge,? and attaches that information to its request to the network operator. Users can go ahead and even specify, for example, the edge server where they want to process data, because they may know that only that specific server can provide the security they expect. Monica: Within ETSI, there is an industry-wide effort to standardize edge computing with MEC. A MEC standard is an enabler for edge computing in multi-vendor networks. You need multiple vendors for edge computing, and interoperability is required. What is the progress to date? Debashish: Standardization is important for MEC technology to be adopted widely. There are areas REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |60| where standardization is required, and some that we can leave open. For example, MEC is a combination of two technologies. One is the networking technology, and the other is IT, or information technology. From the networking technology perspective, a MEC platform has to be connected with the core network and with the access network. We may need to standardize that interface; otherwise we may have issues with interoperability. Also, we may need to standardize how a MEC platform will collect and provide network information to applications. The information that the platform provides may be a candidate for standardization. If we look at the IT side of the MEC technology, we may need to standardize how an application can be run on a platform, what APIs it can use, and what services the platform provides. By standardizing those APIs, we enable developers to write applications for platforms from different vendors. That is also a very important part of the standardization. But we can leave open certain decisions about implementation, such as how the application provider reads, processes and uses the data. Those are definitely aspects that we do not plan to standardize. Monica: The standardization process is going to create a foundation for an ecosystem for MEC and other edge-computing initiatives. What is InterDigital doing to strengthen this ecosystem? Debashish: We believe the creation of the ecosystem is important. This can be done in multiple ways. First, by participating in the standardization efforts. Second, come up with test beds and proofs of concept. At InterDigital, we have developed a FLIPS-based proof of concept available in the ETSI PoC Zone. We also actively participate in many of these consortiums, where we contribute towards building test beds to evaluate the MEC technology. ETSI MEC and other consortiums are developing these APIs that allow application developers to come in and build applications for the platform. In ETSI MEC?s phase one work, APIs have been defined already. They will enable application developers to build and run their applications. Many vendors are coming up with platforms that run applications at the edge. We believe that in the near future, with the standardization of APIs and the development of test beds and proofs of concept, we will see more development and growth in the MEC ecosystem. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |61| Monica: What is the relationship between 5G and MEC? Do we need to wait for 5G before we deploy MEC? Debashish: We see a lot of references about the link from MEC to 5G, and from 5G to MEC. Today, MEC is being deployed in 4G networks, to offer innovative services and enhance user experience. However, MEC will hit its potential with 5G, and 5G will hit its potential with MEC. They complement each other. MEC is the very foundational element that will allow 5G to live up to its potential. In the IMT-2020 5G use-cases triangle, MEC is a foundational element in a lot of the cases in the ultra-reliable and low-latency vertex. For example, the requirements for use cases such as connected gaming, cognitive assistance, industry automation, mission-critical applications cannot be met without MEC. Furthermore, MEC provides network-related information that will be valuable in 5G networks. Monica: Virtualization and MEC are closely related. It?s difficult to think of mobile edge computing without some level of virtualization, because virtualization allows you to move functions to the edge, or it makes it easier. Debashish: MEC is a combination of networking technology and information technology, otherwise known as cloud technology. If we talk about cloud virtualization, use of virtual machines is what comes to mind automatically. The ETSI MEC reference architecture describes the virtualization of the MEC platform. ETSI MEC actively collaborates with another ETSI group called ETSI NFV, involved in Network Functions Virtualization. To enable cloud resources at the edge, we definitely need virtualization techniques. It?s not only the VM, or the virtual machine, technology that will enable MEC or make MEC successful. We need to also look into the other technologies that are available. One example of this is containers, or micro services. They enable virtualization, but they?re different from a virtual machine concept. Those will also come into the picture as MEC gets deployed into multiple use cases. Monica: There are different definitions of edge computing. Where is the edge? You can argue that the edge is at the eNodeB, or it could be an aggregation point in the C-RAN. Or it could be in the device. Where is the edge? And is there one edge? Debashish: We think the edge can be at multiple levels of the network. Again, it really depends on the use case, or the verticals that we are talking about. Now, if we talk about, say, ultra-reliable low- latency use cases, it makes sense for the edge to be defined as what we call the leaf of the network, the extreme edge. It?s not even eNodeB. It may not even be a small cell. It can be the user device itself. In certain cases, where ultra-low-latency is not required but, say, we need a big amount of computing, it?s OK to put the edge at one or two levels higher in the network. Maybe at the core network level, or in certain cases, it?s OK to be in the distant cloud itself. It?s based on the requirements of the verticals we are considering. Where we want to apply the MEC technology will decide where the edge is. From an application-developer perspective, before writing down the code or developing the application, a developer needs to first clearly identify where the application will benefit, and decide where it should be deployed to get the most benefit out of the edge computing. Based on those requirements, the application will be deployed at a certain location in the network. Monica: What are the best use cases for mobile edge computing, and where should they be located within the edge? Debashish: For the ultra-reliable low-latency communications, or URLLC, use cases, we see applications like connected gaming, cognitive assistance, autonomous navigation for self-driving cars, or even drones benefiting the most from edge computing. It would be beneficial for those use cases, where we need very low latency, to move computing to the edge of the network, what we also call the far edge ? maybe in the small cells, or maybe in the access points. Or even certain parts of the edge resources can be put in the device itself. In other cases ? for example, photo editing or OCR applications ? computation may not need to be in the device itself. We can offload computation to the network ? say, one level higher in the aggregation, in the eNodeB, for example ? or maybe at the core network level. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |62| Monica: Let?s look at the business model. Are operators going to pay for MEC, or are the content providers, the enterprise, the venue owners going to pay for some or all of MEC?s infrastructure? Debashish: The business model is still evolving. The industry doesn?t have a clear answer to questions like ?What?s the business model?? or ?Who pays for MEC?? We think the initial adoption of MEC will be driven by enterprise customers to support some of their applications, such as industrial IoT, surveillance, covering sporting events, distributing content generated from those events. Now, moving beyond enterprise customers, there are use cases where both end users and the network operators benefit, in which case the question of ?Who pays for that?? is still not clearly answered. Network operators may deploy MEC with the view that they will be able to generate additional revenues. Revenues may come from the users or the application developers. By providing services such as computation, network information, location information, etc., network operators are hoping to generate additional revenues. On the other side, users may not be willing to pay, because, from their point of view, operators deploy edge computing to solve a network-related problem. That is an area which is still not clearly defined, but we think it will be driven by how successfully MEC platforms solve problems related to latency, backhaul traffic and computational offloading. If MEC enables innovative services and improves the performance of applications, users may be willing to pay for those value-added services. Monica: Can you tell us what you?re working on right now in preparation for the next five years? Debashish: At InterDigital, we are still debating exactly where the edge is. Is it at the core network, or is it at the eNodeBs, or is it in the devices? From our perspective, we think that we will benefit by pushing applications to the far-off edge ? i.e., very far out in the network. We are interested in enabling applications that will require this far-edge MEC to support low-latency use cases. We are now working to build a MEC platform that will enable running in far-edge devices. When I say ?far-edge devices,? I mean a small cell or a Wi-Fi access point, which will not need, say, any additional computational infrastructure. We will try to use whatever computational capability those devices have. We are planning to build a platform that can be used across all these different devices, and enable application developers to write one single application for multiple devices. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |63| About InterDigital InterDigital develops technologies that are at the core of mobile devices, networks, and services worldwide. We solve many of the industry's most critical and complex technical challenges, inventing solutions for more efficient broadband networks and a richer multimedia experience years ahead of market deployment. InterDigital has licenses and strategic relationships with many of the world's leading wireless companies. Founded in 1972, InterDigital is listed on NASDAQ and is included in the S&P MidCap 400? index. About Debashish Purkayastha Debashish Purkayastha is a Member of the Technical Staff in the Technology Evaluation and Prototyping Group at InterDigital. He is part of a team working on Multiple-Access Edge Computing (MEC) and currently focusing on activities related to MEC in 5G and multi RAT environments, and to distributed virtualization to enable computing at the extreme edge of the network. He has been working in the wireless communications industry for more than twenty years, focusing on the design and development of 3G, 4G, 5G cellular and Wi-Fi systems. He has been granted 25 patents and numerous patent applications are pending to date. He holds a Master?s degree in computer engineering from Villanova University in Villanova, Pennsylvania, USA. . REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |64| Profile Qwilt Qwilt was founded in 2010 to help broadband fixed and wireless service providers optimize the delivery of video traffic, both to meet the capacity and latency requirements of high video traffic loads, and to improve quality of experience. From the beginning, Qwilt has developed solutions based on open caching to manage video content from multiple sources and direct it to a variety of subscriber devices that work across different types of networks. Depending on the application requirements and service provider preferences, Qwilt solutions can be deployed in the core, at the Gi-LAN, or in the edge cloud, at the eNodeB. Qwilt?s Open Edge Cloud platform provides open caching and content delivery solutions for service providers that want to optimize video delivery and other real-time applications such as augmented reality and virtual reality at the edge of the network. The Edge Cloud platform leverages compute and storage capabilities as close as possible to the edge ? and hence to the users ? to minimize latency. It relies on cloud management and connectivity, and open APIs using Edge Cloud Nodes. The Edge Cloud platform is not designed to replace or compete with CDNs, but rather to complement them, and carry the traffic where CDN nodes are not deployed or are not cost effective. Qwilt?s Edge Cloud solution aims to extend the content providers and CDNs footprint, to lower transport costs and to improve delivery quality. It also allows them to manage traffic spikes and adjust to the uneven distribution of traffic through time, and, hence, to utilize network infrastructure more efficiently and reduce the need for capacity expansion. The solution allows service providers to meet anticipated traffic spikes (e.g., a game or an update) or unexpected ones (e.g., an accident or a viral video). It can lower capex (less infrastructure investment is needed) and lower opex (peering and transit costs are lower). Open Caching software combines open caching with media delivery and analytics, and runs on an NFV platform, the Video Fabric Controller, that runs on COTS. With open caching, data that is used frequently is stored at the edge and delivered when requested, without requiring any action from content providers, CDNs, or subscribers. Qwilt estimates that 10% of titles account for over 80% of video traffic. Without caching, this frequently accessed content has to repeatedly traverse the network, increasing costs and latency, and lowering QoE. Caching at the edge enables operators to reduce traffic and cost within the core, and minimize latency. According to Qwilt, with open caching operators see an average streamed bit-rate increase ranging from 55% in North America, to 133% in Asia. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |65| Qwilt Video at the edge optimizes fixed and mobile networks A conversation with Dan Sahar, Co- Founder and VP of Product Marketing, Qwilt Monica Paolini: Video is the traffic type that benefits the most from edge processing and storage, both in fixed and mobile networks. And increasingly they both share platforms and traffic management tools. Edge computing will accelerate this trend. We explore these topics with Dan Sahar, Co-founder and VP of Product Marketing at Qwilt. Dan, can you tell us what is that you do at Qwilt to help service providers improve the subscriber experience with streaming video? Dan Sahar: Qwilt was founded in 2010, with the objective of helping service providers address the growth in online video. Streaming video was just starting back then, with the likes of Netflix and YouTube, and that had a major impact on service provider networks. We realized the best way to address that challenge was by changing the way broadband networks and mobile networks are built, and by bringing the content delivery function of those videos, primarily, closer to the subscriber. And by doing that, we gain efficiencies on multiple fronts, both on the economic front as well as on the quality of experience front. The solution we created is one that enables these service providers to do exactly that. You could think of it as the last tier of content delivery that sits inside the service provider network, and is able to acquire content from a range of sources, and deliver it in close proximity to where the users are. Monica: That?s a major challenge, because video is the most difficult type of content to transfer over fixed or mobile networks. Can you tell us what you do to the content itself? How did it change throughout the years? Dan: One key principle in our solution from day one was that we maintain the same fidelity for the video as it was originally streamed and thus we do not make any changes to the videos themselves. Video has changed in several ways through the years, from progressive download initially, into adaptive bit rate. Adaptive bit rate is probably the method most streaming video providers use today; progressive download has pretty much faded out. And the move we?re seeing right now is to use adaptive bit rate for both live and VOD content, and to look at ways to optimize and secure that delivery. We?re also seeing a growth in TLS and HTTPS delivery mediums. Our solution evolved to address ABR, as well as changes on the transport and content sides. Monica: When we talk about edge computing, how do you define the edge? Where is the location of the edge that optimizes delivery of video and other types of content? Dan: We see the edge in primarily two locations. Inside the network, it would be the first IP location in the network. On the fixed-line side, like in a cable network, that might be a CMTS location, and it can be on the B-RAS on the fixed-line side. On the mobile side, it used to be at the SGi or Gi level. We?re seeing it move deeper, to the eNodeB and S1 interface. That would be the first point of edge inside the network. The second place where the edge can play a role is on the device or at the home. On the device, this can be the handset and the software application running on the device. At the home, it can be a residential gateway, or even an Apple TV or a Chromecast or an Amazon Echo that has some built-in content delivery capability inside of it. And there?re different characteristics for each one of these locations: some have more processing power and more storage; others have less processing power, less storage, but they?re a lot closer to the consumer, so they bring more value to the entire value chain. Monica: How do you decide where the edge is for a specific type of content, application, content provider or service provider? Is there an easy way to figure it out? Dan: I think there?s no one right answer for that. You could equate this to the way packages are delivered in the real world. You can have US Postal Service, you can have FedEx, pretty soon you?ll have drones delivering them to your home. They all get the package there on time, but some of them cost more, and each one has different capabilities. That?s a good analogy to how content delivery is done. Some things you can deliver from the centralized cloud ? for example, from an AWS data center ? and for other things, there?s a lot more value to doing them at the deep edge of the network. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |66| Ultimately it has to have an economic balance. There has to be a benefit to the service provider to deploy edge computing and storage resources so that content delivery can leverage these capabilities. Then you have to decide which content can be delivered from the centralized cloud, and which content has to come from the edge. Monica: At the beginning, video was over fixed networks. And when video over mobile came along, it was a completely different thing ? just short videos, and different from video over fixed networks. Is that still the case, that we can make a clear distinction between fixed and mobile? Dan: I think they?re becoming very, very much the same. And operators are changing as video changes. On the consumer side, people watch videos on their mobile devices. It can be on Wi-Fi, but when they get out of their own homes, they continue to watch the same videos. The content is becoming a lot more similar across delivery mediums. Maybe you have a bigger screen on one than on the other, but the content is not different. There?s some adaptation the content provider has to do for the screen size, but, other than that, the transport medium and the ABR formats are exactly the same on both access types. And I think it?s a good thing: the industry is becoming one big video medium, so you can watch video wherever you are. Monica: And the expectations are pretty much the same from the user perspective, so they?re not willing to say, ?Well, since it?s mobile, it might not be as good quality.? They expect the same good quality they expect from fixed networks. What does this convergence mean for Qwilt in terms of the solutions you provide? Dan: Our solution has two main components. One is the edge-cloud nodes, the cache software that sits inside the service provider network. The second is the cloud component ? you can think of it as the control plane ? decides how to delegate the traffic into those caches. Now, these edge cloud nodes can sit on the mobile side, and they can sit on the fixed-line side. What goes into them has to be location specific. If I have a software node that sits on the fixed-line side, it will cache the content that is relevant for that part REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |67| of the network. If an operator has both mobile and fixed, maybe it?ll have nodes that have the mobile formats for those videos on the mobile side, while the equivalent nodes on the fixed-line side will have encoding that is more suitable for the residential devices such as Apple TV or Chromecast. But their function will be very much the same. Monica: That?s good also for those service providers that have both fixed and mobile. They can manage video and other content using the same platform on both networks. Dan: The operator will have a single platform it can manage across both mobile and fixed, and this platform will also be able to address the operator?s own content. Many operators have a video side to their business, as well as third-party OTT content that comes over the internet. The vision is that there will be one layer that addresses both of these content sources, and there would be a single resource that can adapt and handle both. If an SP has a big launch of a new series, you can allocate the resource to that. If there?s a big live event going out from OTT sources, the resources will shift to that. Monica: You raise an important issue: both the service provider or a content provider may own the content. As we move more functionality to the edge, how is the relationship between content provider and service provider changing? Dan: With our solution, we help the service provider become part of the content delivery chain. We do this by creating an API that enables various content providers to make use of these resources. But the owner of the actual components ? the storage and compute ? is the service provider, and the service providers have to decide how to make use of these resources. We give them the tools to do exactly that. I think the understanding across the board is that consumers expect to have content from a range of sources ? not just from the service provider?s own offering ? and the service providers have to build the network that can handle them all. Monica: The service providers are in control of how the content is transmitted, but at the same time, they need to work with anybody that provides that content to make sure the content delivery is working as expected. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |68| Dan: Exactly. I wouldn?t say take control; I would say they participate in content delivery. Because the current delivery medium is still valuable and is still going to be used. A lot of content is going to be streamed from the centralized cloud platform. For some things, such as popular content, or content that is very latency sensitive, you can put into play the edge resources that the service provider has deployed. These are network resources that nobody else has, and content providers can decide which content to stream from the centralized cloud, which to delegate to the service provider for delivery at the edge. I think operators could play a much bigger role than they do today in content delivery with edge cloud delivery. Monica: Within edge computing how much are efforts such as MEC and Fog Computing accelerating this trend? Dan: There are technical challenges on several fronts. On the mobile side, one is how to get a server function to be inside the mobile network, and MEC is doing a lot of work on that front. If you?re deploying content delivery functions on the S1 interface, how are you going to take care of billing? How are you going to get into the GTP tunnels? There is a number of questions. You could think of them as inter-working function questions covering how to place a server inside a mobile network. And I think the MEC group is doing a great job on that front. There?s another set of interfaces that have to be defined: how does a content provider that has a VOD or live offering on the internet make use of these service provider resources? How does a SP publish these resources to the outside world as a function that content providers can leverage? Qwilt has been doing a lot of work on that front under the scope of an industry group, the Streaming Video Alliance, which is creating the set of APIs to enable a content provider or a commercial CDN to use what?s known as open caching functions that sit inside the service provider edge cloud. Monica: You?ve been working on this before MEC standardization work started. What have you learned so far? Dan: The ecosystem needs a lot of balancing. You have a situation because some service providers compete to some extent with internet content providers ? they both have video services ? but there is the common understanding that the consumer is shared between the two. And there?s a greater maturity in the industry now: people are figuring out that the content providers, CDNs and service providers have to work together to create an infrastructure that will benefit everybody economically, but also quality-wise. That?s something that took many years to build, and when we started out there was little collaboration between the two sides ? the service providers and the content providers. It?s becoming a lot better now. Monica: And I think that?s crucial, because, as you said in the beginning, monetization is a big issue. Is somebody getting a free ride? And do you think there is now a balance within the ecosystem as to how much different players benefit and have to commit to financially? Dan: You see several initiatives that are driving more collaboration, like the Streaming Video Alliance that I mentioned earlier. The TIP project that Facebook is driving is another such initiative. And ultimately you see a far greater collaboration between the two sides that didn?t exist before. Service providers have a lot to offer to the ecosystem, and they bring important assets. They own the network, and the network has capabilities that are unique, that you cannot just get anywhere. If I?m an internet content provider, there?s no way I can get my hands on multicast in the access network without a service provider to help me out. And then on the other side you see a service provider understanding that content providers drive a lot more data into the network, which is good, and ultimately, that?s what the consumers consume ? a lot more over-the-top content ? so their network has to support that. Monica: What does Qwilt do that is crucially needed and different from everybody else? Dan: The first fundamental principle at Qwilt was that we are at the edge. We?ve been deploying software at the edge of the network where compute and storage resources did not previously exist, and we were one of the first companies to do that. And the edge has a lot of intricacies that are not trivial to solve. How do you manage massively distributed software nodes? How do you squeeze the most performance out of the very limited computing storage you have at the edge, because of the real estate limitations? The other aspect is the ecosystem that we?re building, both technically and commercially, with REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |69| content providers as well as CDNs. We?re ahead of the curve in terms of the APIs that are required to drive this collaboration ? how content providers and CDNs can make use of the edge cloud of the open cache function that sits inside it. Monica: The business case is becoming a hot topic in edge computing. With edge computing, you obviously need to add more infrastructure at the edge, and that comes with a cost. You get better performance, but is this an appealing tradeoff? Dan: You have to look at who?s gaining what from this edge computing model. Every bit that is streamed from the edge inside of a service provider?s network is a bit that doesn?t have to cross the entire network, and it doesn?t have to cross the core of the internet. That means a service provider gains economic efficiency by streaming that bit from the edge instead of coming from a transit or peering location. And there is value in that. The other value is consumer experience. Because content is streamed from the edge, there?s a greater ability to overcome any bottlenecks in the network, and much lower latency as the content is streamed to the consumer. That?s a benefit to the service provider as well as to the content provider in terms of the experience their subscribers are getting. If the content provider is striving toward providing a full HD experience all the time without buffering, this is the way to do it. And this value extends to commercial CDNs that are the platform many content providers use to distribute their content. This gives them better reach in places they cannot reach today, and with far lower latency. There is definitely value to the entire ecosystem and a question of how each side is going to compensate the other for this value. I think the market will dictate how exactly that is done. But we have found that equation balances itself out, and when we?re streaming content from the edge, there are economic benefits across the board for the ecosystem. Monica: Do you see, in terms of ownership and initial capex investment, an increasing role for the venue owner, the content providers, the CDNs, to participate, because it?s something they benefit from along with the service provider? Are they willing to step up and put some investment into it, as well? Dan: That?s something that will fall primarily on the shoulders of venue owners and service provider as they?re building their venue network. It?s expected of them, because that?s they will reap the benefit. And for edge computing, it?s like the move from traditional networking into networking that uses storage and compute resources, so it?s basically shifting resources from one place to another place. Both network operators and venue owners are the ones that are going to be responsible for building out that infrastructure, and there are ways for both of them to benefit. Over the long term, that?s a far more economically sensible way to build networks than simply to throw routers at the problem. Monica: What are you focusing your attention on these days to get ready for the challenges, including the challenges of 5G, over the next five years? Dan: Our focus is on two fronts. One is the technological front: to create standardization around the APIs that are required to enable this open cache function inside the edge cloud, and to manifest that into our product as well, and to have a range of APIs that the content ecosystem can use ? increase our capabilities when it comes to service providers? own content. The second front is on the ecosystem side. We?re trying to enhance our content provider and CDN relationships so they can make use of this edge cloud layer that sits deep inside the service provider?s network. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |70| About Qwilt Qwilt?s unique Edge Cloud platform and Open Caching software solutions help internet service providers address the dramatic growth of streaming media on their networks and the need for a low latency, high scale infrastructure to support future applications. Qwilt?s cloud-managed open platform, running on commodity compute and storage infrastructure and deployed close to consumers, creates a massively distributed Edge Cloud that supports applications such as Open Caching, 4K live streaming, AR, VR, self-driving cars and IoT. This low-latency Edge Cloud architecture enables a high-quality streaming experience for consumers on a massive scale. A growing number of the world?s leading cable, telco and mobile service providers rely on Qwilt for Edge Cloud applications. Qwilt is a Founding Member of the Streaming Video Alliance and a leader of the Open Caching industry movement. Founded in 2010 by industry veterans from Cisco and Juniper, Qwilt is backed by Accel Partners, Bessemer Venture Partners, Cisco Ventures, Disrupt-ive, Innovation Endeavors, Marker and Redpoint Ventures. Learn more at www.qwilt.com. About Dan Sahar Dan Sahar is the Co-Founder and VP of Product Marketing at Qwilt. Dan drives product marketing and go-to- market activities for Qwilt, bringing more than fifteen years of marketing and product management experience at high technology companies. Prior to co-founding Qwilt, Dan was Director of Product Marketing at Crescendo Networks (F5 Networks), a leading provider of data center application delivery products. At Crescendo Dan was responsible for leading the company?s overall marketing and product direction. Earlier in his career Dan held product management roles at Juniper Networks and Kagoor Networks (acquired by Juniper) as well as engineering management positions at Kagoor Networks and Seabridge (Nokia Siemens Networks). Dan holds a Bachelor?s degree in Computer Science and Business from Tel Aviv University Magna Cum Laude and an MBA (Marketing) from the Leon Recanati School in Tel Aviv University. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |71| Profile Vasona Networks Vasona Networks? focus has always been optimizing RAN performance and QoE by managing traffic at the edge between the RAN and the mobile core, and it started well before MEC and other recent edge computing initiatives were established. The goal of this approach is to help mobile operators improve end-user quality of experience and utilize network resources more efficiently and, in turn, save money and differentiate their services. To date, Vasona has raised $48 million from investors that include Bessemer Venture Partners, New Venture Partners, NexStar Partners and Vodafone. Vasona has deployments in Europe, North America, and Latin America. Vasona?s initial solutions addressed the challenges that mobile operators face with video traffic and its tight capacity and latency requirements, which have to be accommodated in a RAN that is often congested. Bad video experiences not only cause end-user dissatisfaction, they also waste RAN resources. In the last few years, Vasona?s approach has widened to include the optimization of all types of traffic, by allowing operators to manage each type of traffic based on specific traffic requirements, RAN conditions, and operators? strategy and policy. Vasona has developed standards-based software platforms for MEC that sit at an aggregation point between the RAN and the mobile core. By placing the MEC functionality in an aggregation point that, typically, covers a thousand or more cells, Vasona helps operators contain the cost of edge processing and coordinate traffic management over a larger part of operators? footprint, rather than for a single cell. Vasona has two products today. SmartAIR? is an edge application controller designed to overcome resource contention at the individual cell level, taking into account all the active application sessions. When the RAN is congested, SmartAIR operates in real time to manage individual traffic flows to reduce latency and improve network utilization. SmartVISION? is a software suite that provides operators with real-time and historical data to help them analyze RAN performance. For each cell sector, SmartVISION collects information on user activity, app and content usage, and capacity ? information that operators can use to optimize network performance and plan for network expansion. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |72| Vasona Networks The network performance edge A conversation with John Reister, Vice President of Marketing and Product Management, Vasona Networks Monica Paolini: MEC and other edge computing initiatives call our attention to the relevance of location of network functions in determining performance and, even more crucially, QoE. Vasona Networks had started to move traffic optimization closer to the edge before MEC work at ETSI started. In this conversation with John Reister, Vice-President of Marketing and Product Management at Vasona, we talked about the evolution of moving functions toward the edge ? and about where the edge is. John, can you tell us what Vasona does in this space and what?s your role in the company? John Reister: I run product and marketing. Vasona is an edge computing company. We were founded on the principle of finding value for mobile operators by providing computing software at the edge of the network. Monica: Over the last few years, we have been talking about the cloud, about centralizing everything and moving processing away from the edge. In the last few years, there has been an increasing focus on what?s happening at the edge ? and on edge computing. Why do mobile operators care about edge computing? John: One of the things edge computing does is it enables operators to achieve a higher degree of flexibility in their networks. That gives them a lot more agility in how quickly they can move and how quickly they can introduce services. It has the potential to reduce cost. It enables operators to optimize networks and introduce new service capabilities for their customers. Ultimately, it?s about enabling them create more value in a way that?s unique. When operators implement technologies in the core or in the data center, they tend to be things that can be easily copied by anybody on the internet. Whereas when you do things at the network edge, it?s a more sustainable differentiation, a unique capability that adds value for your mobile customers. Monica: Are there any challenges that come with it? John: Yes. One of the challenges, is when you say edge computing, where?s the edge? One of the challenges we hear from the operators is ?I?ve got 10,000, 20,000, or 50,000 cell sites. If you?re going to tell me I have to deploy a server at each one of those, or even one server for every 10 of those, that?s thousands and thousands of servers that have to be deployed and managed. It?s not only expensive, it?s an operational mess.? That?s why we don?t do it that way. Alternatively, we deploy at an aggregation point, where there tends to be anywhere from 100 to 500, or even 1,000, cell sites that come in through that edge compute implementation. You?re two orders of magnitude, almost three orders of magnitude more scalable, because there are fewer places you have to deploy. The other big challenge we encounter is that edge processing is an inline element. To do the kinds of things you want to do with these services, you need to be inline. As soon as you?re inline and affecting tens of thousands, if not hundreds of thousands of mobile customers from this installation, reliability is critical. Achieving high availability, high reliability, is essential. Proving that you can achieve that with this kind of infrastructure is key. Monica: How is Vasona addressing the challenges you just mentioned? John: Our implementation is software based. We can be deployed as close to the edge as the operator wants. But for the applications we do and for the kind of capabilities operators want to achieve with the platform, we think that being at the aggregation point works really well. We?ve implemented capabilities to sit at that point and inline and do the traffic classification. From that point, we can map the users and figure out what cells they?re in. We can then classify and guide the traffic through the applications the operator wants to apply for those users. Monica: You have been working at the network edge since before work on MEC started. How long have you been doing this? REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |73| John: We introduced our product, SmartAIR?, a little over four years ago. We called it an edge application controller, which is a pretty similar name to multi-edge later adopted by ETSI. We?re glad edge computing now has a home in the standards community, with the MEC initiative. We introduced an edge solution to provide unique value from this point at the edge: low latency capabilities that allow operators to manage the challenges they face ? not just the ones on the road to 5G, but the ones they?re currently facing. We?re broadly deployed. We have over 100,000 cells deployed in Europe and the Americas. Monica: You?re beyond the proof-of-concept stage that most of the MEC trials are at. What have you learned so far from your operators as you try to help them? John: I mentioned already that the operational challenges are very important, and so is reliability. To interoperate transparently, you have to be able to insert edge functions into the network transparently, and to install them easily. These things have been really important. We identified the applications that affect operators and that are needed today. They?re struggling to deal with the onslaught of video traffic. A lot of the tools that have been tried have been made obsolete. There?s a great need to deal with all that encrypted traffic, while keeping the mobile customers happy and committed to the operator as their service provider. Monica: Video is a big issue ? if nothing else, because so much of the traffic is video. And there is a role for processing and optimizing video traffic towards the edge. But where is the edge? If you put processing too far out in the edge, it may not be as effective. If it is too centralized, it may also not be effective. How do you find the right balance there? Where is the right place for the edge, in terms of computing? John: We?re not big believers in caching. We think as you get close to the edge, caching is not as valuable. And it?s not helping to address capacity issues with the air interface anyway. But we do think traffic management benefits by locating it at the edge. One of the best MEC use cases is throughput guidance. We do think there are applications that help improve the quality you can get over existing infrastructures. You avoid REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |74| having to run to stay in place, not providing better service while constantly having to throw tons and tons of resources into new cell sites and carriers, and cell splits, and so on. What drives how close to the edge you have to be is the fact that the traffic is largely encrypted. To figure out that you?re dealing with a video session and what its needs are, you have to see the beginning of the session being set up. If you are too close to the edge, when a user does a handover to another cell, the MEC application on that other cell would not know what that session was. So the application would lose context, and you would have this very complex need to hand over from MEC instance to MEC instance. It gets very complex. By being located at an aggregation point, like we are, where we?re dealing with several hundred cell sites, the vast majority of handovers we see are between cells and between cell sectors that are covered under the same MEC instance. When users hand over from one cell to another, we already know what the conditions are on the cell they?re going to before the very first data packet goes through that new cell. We see the handover from the control messages. So even before the data starts moving over to that new cell, we know they?re going to be in that new cell. We already know the conditions of that new cell, and we can adjust and take that into account in our traffic management, essentially before it happens. Monica: This is an entirely new way for operators to manage traffic. For video, it?s very important. But are other types of traffic affected as well? John: One of the benefits of managing traffic as we do at the application-type level is that you can manage down the latency for those time-sensitive services where the user is directly interacting with the network. By managing traffic at the application-type level, you can keep the queues much shallower. Because of that, the latency you see during the busy hour can be cut by 25% or 30%. That affects browsing. It affects your social media applications. The benefits are not just for video. They?re definitely for all types of services and, obviously, for the coming IoT services. Monica: It also gives operators a better view of what goes on in their networks. John: A part of edge computing that is less often mentioned, is that you?re now linking together not just information about the users and the type of applications they?re using, but also what cells they?re in, how busy those cells are, how congested those cells are. You?re able to put all those pieces of data together. One of the prime benefits is you can now manage your capital investment with a focus on the quality of experience that you?re delivering for those applications that are important to your customer base. Instead of just looking at low-layer metrics of the network, you can now actually manage your capex that way. That allows you substantial savings, and it?s a big contributor to the business case. Monica: Business case, you said it. Let?s talk a little about that. As you said before, you want to avoid the risk of having servers at every cell site. But you still need to add some hardware. This is something that you add on top of what operators already have, so that?s an additional cost. Is there a business case that justifies the additional spending that edge computing requires? John: I think this is one of the keys. You have a MEC platform, and then you?re going to layer your applications on top of it. There?s certainly the potential for new revenue streams. Some are relatively far-flung, and some potentially closer in. But as you said, you can talk virtualization as long as you want, but the historical cost savings of virtualization come from a massive server farm in your data center, replacing all these purpose- directed servers that were underutilized. That?s not the case with MEC, because it?s a new investment. The business case still has to prove itself. There are benefits on the revenue side, from new services. We see the benefits of improving quality of experience: getting more customer advocates who believe in your network because of the reduced stalls and lower latency on the browsing and the social media and so on. But, clearly, the biggie in the near term is investment savings. You don?t have to constantly run out and chase after the peak demand in your network. The problem is, human nature is such that people remember the bad experiences much more than the good experiences. It only takes 5 or 10 bad experiences a week for users to think you?re a lousy service provider. That can easily happen in those peak hours. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |75| You?ve been in this mode of having to build to those peaks, and it?s really expensive. By putting in a mobile edge compute system, you?re actively managing the traffic congestion and managing the services. Because of that, you can make the quality so much better during those peak times. That allows you to not have to use your capex to chase after those instantaneous spikes in usage that are causing those bad experiences. That capital benefit is certainly a key part of the business case ? being able to focus your capital on the portion of the network causing the poor QoE. Monica: The business case becomes subtler here. It?s not just adding numbers, but seeing what the resources you have can do and how happy your subscribers are. It requires a different conceptual framework behind the financial model. John: It does. I think there is a transition in thinking that is starting to happen, because everybody looks at the growth in demand and they look at the revenue curve, and those lines are not parallel on any chart. I think there?s an awareness that you can?t continue to spend to meet that level of demand without something radically changing. This does require new thinking and an ability to focus your capital investment to maximize the quality of experience. That?s one of the things MEC can enable operators to do. Monica: With edge computing, content providers may take a more active role. They might get involved in rolling out or participating in some way with the MEC infrastructure, because they stand to benefit from it. John: This may be the best part, to be honest. For a long time, I think the industry has had a model that?s been ? I won?t call it hostile, but difficult. There has been a situation where the applications create all these demands on the network. With edge computing, you have an opportunity for a truly collaborative approach. You can take IoT. You can take throughput guidance. There are clear cases where you?re now essentially having levers to pull or information in the network that is being shared in real time, and that gives the ability to evolve that relationship between operators and content providers. With throughput guidance, for example, the operator is telling the video streaming content provider in real time, ?This is the best this user can get at this time, and let?s jump right to that rate. Don?t try to go higher, because it?s just going to get stuck and congest the network, and you don?t need to go lower and harm the quality.? You share that information, and it?s best for the mobile customer, who is really the customer of both that content provider and the mobile operator. IoT is another example. You have the ability to provide much better security and lower latency, by bringing that IoT traffic around the core through a VPN into a private cloud. You can create a secure capability with low latency for those types of services. You?re providing levers, information and tools that allow this tremendous, positive working relationship between the service provider and the content provider. Monica: I guess that might address some of the issues we have with encryption. If there is collaboration, then encryption no longer creates a barrier to optimizing around traffic. John: Encryption is a tricky subject. Of course, there are two layers of encryption. Operators put on their own encryption when the backhaul is over a shared environment or non-owned environment. They run IPsec tunnels from the NodeB site back to the network. That certainly affects where and how mobile edge computing gets deployed. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |76| Then you have the encryption that the content provider puts on its content. I don?t expect that to change. I think end users want to ensure privacy and so on. At Vasona, we have the ability to classify and understand encrypted traffic. We don?t decrypt it in any way. But we?re able to figure out what type it is and then make sure it gets the treatment it needs. If there are situations where you are going to host it on the MEC platform, which would be the case for IoT gateways, for example, then you?re absolutely right. The point of the certificate exchange in that case can move right onto the MEC platform for that. Monica: Now, let me ask you the obligatory question, which is about 5G. There are different schools of thought here. MEC is helping the move towards 5G, but is it required? Do we need to wait for 5G in order to use MEC, or is it the other way around? What is enabling what? John: Yeah, that?s a great question. 5G is about a lot more than just the radio. There are huge implications for the backhaul when you get new 5G radios that have much shorter range. You?re going to need more of them. The backhaul gets a lot bigger and more complex. MEC is really a great step in that direction. It?s bringing down the latency. It?s putting the kind of capabilities into the network that you?re aiming for with 5G. It?s a stepping stone. The nice thing about MEC is that it has a business case and addresses today?s problems, while taking you in the direction you want to go. Monica: Let?s have a peek at what?s going on in the next five years. What are you guys working on? John: Obviously, MEC standards are not finished yet. Right now, we?re an edge compute platform. We?re embracing the MEC standards as they move along. There are MEC platform capabilities, and there are MEC applications. We?re working on new applications. IoT is one area. We continue to believe that security and quality of experience remain a big focus. Then on the platform side, it?s integrating down into the network infrastructure, so you can run on a new server cluster. But there are, obviously, other initiatives going on with the packet core and with cloud RAN, that are also NFV driven. You can certainly start to see intersections as you get far enough out. Monica: It looks like you have a pretty busy schedule over the next few years. John: I think it?s an exciting time. I imagine it?s a little bit of a scary time but also a great opportunity for the operators. The demands being placed on them are enormous. The expectations are very high. But then, they?re in the driver?s seat. And that?s a good place to be. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |77| About Vasona Networks Mobile operators work with Vasona Networks to improve user experiences and better leverage network investments. The company?s pioneering SmartAIR? edge application controller and SmartVISION? analysis suite help operators overcome congestion on 3G, 4G networks, build a bridge to 5G, and understand network activities for improved management and planning. Vasona?s standards-based software at the mobile edge delivers intelligent solutions to increase the quality of experience, enable low-latency services and focus investments to create a more flexible, intelligent and responsive mobile network from the individual cell level. Founded in 2010, Vasona has deployments in major networks around the world. For more information, visit www.vasonanetworks.com or contact info@vasonanetworks.com. About John Reister John Reister is VP of Marketing and Product Management for Vasona Networks, supporting the company?s work with global mobile network operators to deliver better subscriber experiences. John was VP product strategy for Arris (Nasdaq: ARRS), joining through its acquisition of BigBand Networks where he was VP advanced technology and chief architect. He was instrumental in the company's expansion to telecom markets with platforms for advanced video services. Previously, John was CTO of DSL pioneer Copper Mountain Networks (Nasdaq: ARRS), a consultant with Bain & Co. and an engineer with McDonnell Douglas (NYSE: BA). REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |78| III. Service provider interviews REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |79| BT MEC in the fixed/mobile converged network A conversation with Mansoor Hanif, Director of Converged Networks and Innovation, BT Monica Paolini: MEC is more than a tool to lower latency. It is an enabler for services that are specific to a location but are served over multiple access networks, both fixed and mobile. In this context, MEC is a powerful driver that accelerates the convergence of fixed and mobile networks. In this conversation, Mansoor Hanif, Director of Converged Networks and Innovation at BT, tells us how this is shaping the network evolution at BT in the UK. Mansoor, what is your new role at BT, and what are you working on with regard to MEC? Mansoor Hanif: Personally, I came from the EE side. I was in charge of the radio networks. Since the beginning of November 2016, I?ve moved to the BT research team. I?m now looking after the converged networks research lab. On edge computing as a whole, at BT we are not in the commercial phase at the moment. We are doing a number of proofs of concept looking at the customer experience ? and especially looking at the converged customer experience that we can manage through the edge and the business cases around this. We did a number of proof-of-concept trials last year, mainly in Wembley Stadium. This year, we have a couple of new locations we?re looking at for proofs of concept. Monica: Can you tell us more about these proofs of concept? What kind of applications, and what were the learned lessons? Mansoor: From the couple of proofs of concept we did last year, we showed the effectiveness of mobile edge computing solutions for reducing latency for video. For enhanced video orchestration in a stadium such as Wembley, we showed that this could be effectively used to replace closed-circuit TV for security uses. There are similar opportunities for many other vertical markets. We also did a throughput-guidance trial with Akamai, and we showed that there is benefit in transmitting the real-time radio conditions through MEC to the CDN; it helps content delivery partners improve the end-to-end experience for mobile video users. Those were successful proofs of concept. This year, we are focusing on a couple of areas. We are having a look at our retail estate to see how we can improve the engagement of our customers when they come to visit our shops. Can we showcase the offers we have and the converged capabilities we have? Can we improve the user experience by localizing some functionality from our IT systems in the central back office to the local shop? We?re going to have a look at that from the retail shop attendant and customer perspective We are also looking into public spaces such as museums, where we can improve the user experience of visitors through edge computing. For the entertainment field, we?re looking into where edge computing can improve the user experience with things like augmented reality. There, we look at local positioning and at low- latency applications, and how we can hook location-based augmented reality applications into our fixed offerings to offer enhanced solutions people can take away and keep with them after they?ve visited a location. We?re also looking at the scenario in which you have a lot of people visiting a location who are from outside the country or from other operators. Can we have a single solution that manages all the users at a location, but also has intelligent control and enhanced added-value services over the MEC platform? Monica: When I started working on MEC, video seemed to be the major use case. What I?m hearing from different sources now is that there is a shift. Video is still one of the main use cases, but there is more to MEC than video. Would you agree with that? Mansoor: Certainly. Anything that requires compute capability to give an effective user experience is a good use case for MEC. Anything that requires a lot of processing power to be able to deliver a good-quality experience. Augmented reality is a good example. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |80| Video, certainly, is a good use case, especially high- definition 360 panoramic video that you want to make available to the user to browse through and choose whatever camera angle he wants. Rather than sending all of that data over to the user, you can do the computing on the edge platform, delivering a very good quality of service to the user, and not expose your network. Monica: Mobile edge computing clearly relies on network virtualization. Initially, virtualization efforts were focused on moving everything to a centralized location ? the data center, the cloud ? to get cost savings and have everything in one place, which is attractive. With mobile edge computing, the trend is in the opposite direction. There are some things that do not work sufficiently well in a centralized location. There are advantages to moving processing and storage to the edge. Mansoor: Today, as a converged operator, we would rather speak about the converged edge for fixed and mobile. One of our key objectives is to make sure we effectively implement convergence across fixed and mobile. If you?re talking about the fixed network and the mobile network separately, the edge can mean different things. There are three elements to this, but there?s probably some granularity between them. First, there?s the capability you would like to centralize into the core. That?s very much about cloud computing and cloud connectivity, where you can effectively centralize. I think that will continue, to some extent, but not to the extent that we have it today. Second, there is the edge seen from a fixed network, which is pretty much on the distributed exchanges or aggregation points. There are some applications we feel should be decentralized there. Finally, there?s an opportunity now to offer converged mobile and fixed aggregated capability on those aggregation points. There?s also still, we believe, a big opportunity at the edge of the radio network, which is the closest to the customer. From our perspective, all of that can fit into an interesting business model: you can leverage a single platform that allows you to manage all those applications anywhere between the core, the aggregation point, and the radio edge with a single platform. That makes it very easy to manage and shift applications when you need to, where you need to. It?s not manageable, not possible, to get to a good business model with completely separate platforms for the radio edge, the aggregation points for the fixed edge, and the core. Two factors are driving things out to the edge. One is that the edge speeds are moving so fast that the kind of capability we need on the edge is getting increasingly difficult to aggregate in the center. Speed at the edge is scaling up so very fast that I don?t think it?s possible to scale a centralized computing platform to aggregate all of this on the long term The second factor is that the processing power required to support those edge speeds is such that you will almost certainly have to enhance the processing power at the edge of the network to some extent, so the utopia of virtualizing all of the radio baseband hardware is unlikely to happen everywhere. If you need to add or upgrade hardware to support the new radio capabilities and speeds, you can also put in some mobile edge computing power and add some intelligence there. In the UK, we?ll probably be doing a lot of indoor installations over the next three to five years. If you are going to do that, then the actual cost of putting in an extra board or two to enable edge computing is affordable. The overall cost of adding the MEC hardware as part of a wider indoor coverage installation is much lower than if you are doing a dedicated MEC rollout. Monica: You raised a lot of very interesting points that I would like to follow up on. The first one is the fixed/mobile convergence. In fact, it?s telling that MEC doesn?t stand anymore for Mobile Edge Computing, but for Multiple-access Edge Computing. What does that mean to you? A mobile network was very different from a fixed network, in terms of what subscribers do. The convergence is not just in terms of technology, but also in terms of what we do over a fixed/mobile network ? or a wireless/wireline network. Mansoor: You said two things there that are slightly different. There?s wireless and wireline; in that case, Wi-Fi would be considered wireless. It?s mobile and fixed, where Wi-Fi would be REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |81| considered part of a fixed network. That?s generally how operators have seen it in the past. Now we?re trying to converge all that into a single, coherent control point. We?d like to have the ability to improve the customer experience as you move between those. We need to make sure we get the best out of our Wi-Fi investment and our 4G investment, and future 5G, we hope, and fixed line. Today, the mechanisms available for controlling those are not sufficiently aligned between the fixed standards and the mobile standards. For example, we need to get a single identity, a single authentication. We also need effective anchoring and control mechanisms to allow us to do intelligent traffic steering between the fixed and the mobile. To this end, we are working with the standards bodies to make sure that, as we move into 5G, we have a lot more ability to intelligently steer the user into the best customer experience over whatever access type is most suitable. Edge computing plays a role there. We believe that on the network-wide level, some functionality needs to stay in the core network. But some core functions for specific applications could also be localized into a MEC solution where, in a specific environment like a museum or a shopping center, you can put in an anchor MEC-based solution that?s hooking into the indoor installation for Wi-Fi and 4G small. Solutions like MulteFire, or LTE unlicensed, or simply Wi-Fi can be integrated into a single MEC anchor, where we can provide layered services. It?s a way of integrating that locally to work together with what?s in the core, but also to work independently and provide extra services where needed in those specific locations. Monica: As an operator, you have to decide which functionality should be centralized and which functionality should be pushed to the edge. And then, for the access network, you have to decide which applications should use the fixed network and which should use the mobile network ? and which fixed or mobile network to use, when multiple ones are available. Mansoor: You put many things on the core side: the unified sign-on, the unified authentication, but especially the traffic steering, the quality of service, and the quality of experience management at the network level. But you can also have a separate policy locally for a specific location, based on what you agreed with landlords and what they want to offer your customers, and other people?s customers. It?s important to have the flexibility to tailor this to local requirements. In the terminology of ?Network slicing?, for example, this is the capability to offer an enhanced network slice in a localized environment. Monica: Do network slicing and edge computing go hand in hand, complementing each other, in the scenarios you describe? Mansoor: Yes. Mobile edge computing adds an extra granularity to the type of slice you can offer. Already, we?re experimenting a lot with how far we can push network slicing on our 4G network. It?s going to be a lot easier with a 5G network, because 5G is built around network slicing. MEC increases the granularity of the type of slicing you can offer because you can then actively offer completely different types of slicing locally for any specific customer. Monica: We talked about the edge, and you mentioned aggregation points and the RAN. Where is the edge? Mansoor: The edge of a fixed network has been considered to be, let?s say, the local exchanges, or the equivalent. The edge from a mobile network is the radio antenna, which is the closest to the customer. Whether that?s a macro site or an indoor site, that?s where the edge would be. That?s where I think the different definitions of the edge have come in. Obviously, depending on the type of application and the type of reach you?d like to have, you can choose where to put the edge capability. What?s important is that it?s also very much dependent on the traffic load and the application load. Ideally, you could have a dynamic capability for orchestration, where you can move an application from the small indoor cell to the macro cell. Or from the macro cell to the local aggregation point or exchange, back and forth, depending on time of day, the load, and the optimal customer experience you want to give. That?s the ideal situation we?d like to get to. Monica: Is this why you need a single platform that allows you to manage dynamically all those elements? REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |82| Mansoor: Yeah, ideally. Otherwise I do not think it?s possible to manage the quality of experience we want to offer customers in order to make it a useful business proposition. It would be very, very difficult to do that. Monica: In terms of the business case, with edge computing, you inevitably have to add some more processing and storage capability at the edge. That?s going to come with some cost. At the same time, it?s clear that the more you push to the edge, the better the performance you have in terms of latency. There?s a tradeoff there. How much is worth pushing to the edge? When you look at the business case, what are the tradeoffs that you think will make sense? How aggressively do you want to push things to the edge, when that comes at a cost? Mansoor: It?s more about being intelligent about the cost. More than a tradeoff, we need to work on all fronts to lower the cost and increase the added value. First of all, we need to get the hardware platforms that support edge computing to be as low cost as possible. That?s why we?re working through initiatives like the Open Compute project and similar ones to get to an almost white box situation for the hardware that supports MEC. We need to lower the cost of the hardware if we want MEC to become deployable on a massive scale. On top of that, the actual cost of implementing a hardware-based solution needs to be reduced. That?s where a standalone business case of rolling out MEC capability into offices and shopping centers doesn?t make any sense from our perspective. However, you can piggyback on indoor installations in an intelligent way so that the increased installation and implementation cost of a MEC solution is only a very small part of your overall cost of the indoor installation. Timing is going to be critical. We need to catch the wave of large-scale indoor installations at the right moment so we can slot in the MEC hardware, at least in the majority of cases. That would change the business model. Those are the two things ? cost of the hardware platform and timing of installation ? that are going to lower the cost of MEC computing on the mobile edge. We need to make sure that when, for example, you?ve got a new customer in a big location, the initial investment is covered by simple use cases of connectivity and some basic services. The MEC API interface needs to be flexible enough that, later, we can very easily add on new functionality as and when we need to, on the same platform. That way, we can continue to generate new revenues on top of the baseline, which is doing the basic financing for the installation. That?s why I think MEC is much more than simply improving the efficiency of the network or using low latency. We need to identify and focus on new services that the combination of proximity and compute power enables, and that we have the flexibility to rapidly implement those solutions on top of the MEC platform as and when customers ask us to. Monica: It?s more than getting lower latency. It?s thinking during network deployment where the functionality goes in a much broader perspective. Mansoor: Absolutely, and being able to dynamically shift content and dynamically deploy new applications locally ? leveraging the value that proximity adds by improving the user experience for applications such as augmented reality. With augmented reality you can be very, very close to the user and therefore really improve the subscriber experience. If you take Pok?mon Go, which was a massive hit, it doesn?t need a network at all, or very little, but it?s not very granular. If you want to take that to the next level and provide services so compelling that businesses are asking us to put them into their locations, you need to make the customer experience a couple of levels better than that. That?s the kind of thing we?re working on so we can offer businesses a compelling way to draw in more customers and have customers pay for more services. We want to enable all of that with a great customer experience that we can monetize, to a certain extent. Monica: When you talked about monetization, you mentioned subscribers. Could you also get revenues from the content owners or the enterprise which also stand t benefit from edge computing as well? Mansoor: I don?t think we would get paid directly by those third parties. But if we come up with the use cases that generate extra revenue for the landlord or the third parties, then we could effectively have a place in that value chain and get a percentage. Monica: What about the capex? Could the mall owner, the stadium owner, the airport REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |83| management be willing to pay for all or some of the infrastructure if it?s for services they provide? For instance, if it?s a mall and it?s trying to get its customers engaged, is this something you think it would pay for? Mansoor: There are already a lot of landlords ready to pay for a good indoor installation, as long as it?s covering all operators and it?s offering a good quality of service. People are moving to that situation. We have a number of companies proposing pre-installation in the UK. It?s good, really. If we have a landlord that?s looking for basic connectivity, we could offer it a shared-cost MEC platform on top of that indoor installation, which would allow the enablement of many new services within that environment. The actual capex of that MEC layer on top of a DAS installation would be only a fragment of the DAS installation itself. It could be partially funded by the landlord in some cases. In my view, if the DAS or the small-cell installation is already being funded by the landlord, it would be reasonable for us to offer to manage a MEC- type service for all operators by putting in the extra capex ourselves and then putting in our added value in terms of use cases we?re offering to the local customers. Monica: We can start with MEC in 4G networks, but with 5G, MEC will be more pervasive and more efficient. How do you see the transition of MEC as we go from 4G to 5G? Mansoor: It?s a very interesting question, because if you are focusing on the latency added-value of MEC alone, then with 5G, MEC?s latency advantage will be taken away, because 5G should be inherently capable of very low latency. On the one hand, you can see how 5G could replace MEC in certain areas. At the same time, to have end-to-end 5G capability, when you?re talking about user speeds of 10G or above for one user, it?s going to be increasingly difficult to centralize all the computing power needed to aggregate that demand. Inherently, if we want to adopt 5G massively, we are going to have to use more distributed computing power, simply because the speeds being offered are so high that it?s going to be very difficult to keep up with the aggregated capacity requirements if you centralize them. With 5G, inherently you?re going to be looking to distribute the core to some extent. Also, I don?t think 5G will be deployed, necessarily, in a very homogeneous fashion in many networks. It will take a few years. In the meantime, you can offer a reasonably homogeneous quality of service across many, many locations by implementing MEC as an enabler in the first place. It will allow us to homogenize customer experience as we roll out services over a mix of 5G and 4.5G and 4G. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |84| About BT BT?s purpose is to use the power of communications to make a better world. It is one of the world?s leading providers of communications services and solutions, serving customers in 180 countries. Its principal activities include the provision of networked IT services globally; local, national and international telecommunications services to its customers for use at home, at work and on the move; broadband, TV and internet products and services; and converged fixed-mobile products and services.BT consists of six customer-facing lines of business: Consumer, EE, Business and Public Sector, Global Services, Wholesale and Ventures, and Openreach. For the year ended 31 March 2016, BT Group?s reported revenue was ?19,042m with reported profit before taxation of ?3,029m. British Telecommunications plc (BT) is a wholly-owned subsidiary of BT Group plc and encompasses virtually all businesses and assets of the BT Group. BT Group plc is listed on stock exchanges in London and New York. About Mansoor Hanif Mansoor joined EE in November 2011 and led the technical launch of the 1st 4G network in the UK and was also accountable for the integration of the legacy 2G and 3G Orange and T-mobile networks. Until 2016 he led the team who plan, design, rollout, optimise and operate all EE radio access networks, including Mobile Backhaul and Small Cells, and was accountable for the coverage aspects of EE?s Emergency Services over LTE programme. He was also a board member of MBNL (the joint venture of EE with H3G) until 2016. During the acquisition of EE by BT, Mansoor led the EE network Integration team and is currently Director for Converged Networks and Innovation in BT R&D. He is a member of the BT Technology Steering Board and is a board member of the Scottish Innovation Programme. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |85| Verizon Wireless Better performance in the enterprise with edge computing A conversation with Matt Montgomery, Director, Wireless Business Group, Verizon Wireless Monica Paolini: Edge computing improves performance and optimizes resource utilization in many use cases. The enterprise is an environment where edge computing is going to play a large role for a diverse set of use cases, which include not only data and voice connectivity, but also IoT applications. I talked with Matt Montgomery, Director of the Wireless Business Group at Verizon Wireless, about how edge computing addresses the connectivity requirements of the enterprise, while providing the same high level of security as centralized services. Matt, can you give us an introduction on your role at Verizon and about what Verizon is doing to bring edge computing to the enterprise. Matt Montgomery: I have business operations, marketing, and partner enablement responsibilities for our Wireless Business Group, which is dedicated to our large and enterprise customer segment. From a mobile edge computing perspective, my primary responsibility is ensuring that our customers can successfully use multiple partnering solutions. It?s not just all Verizon, all Cisco, all Microsoft, or all Apple. It?s a combination of multiple technologies, OEMs, and partnerships to create one solution that solves business challenges. Regarding edge computing, my responsibility is to make sure Verizon provides world-class solutions to customers and that the solutions they choose work seamlessly on the Verizon network. It?s an integrated approach with technologies, equipment manufacturers and partners that create the best solution possible to help them move business forward, faster. Monica: Over the last few years, we have seen a push to move everything to a centralized cloud in large data centers. Now, the tide is turning. Service providers, enterprises, venue owners, and even content providers are showing an interest in moving some functionality to the edge. Why do you think that?s so? Matt: Industry innovation causes a pendulum swing. As new technologies become available, as new threats arise, as the computing experience becomes more intense and form factors become more open, businesses are looking at what they can do locally versus centrally. I don?t believe it?s a binary equation, but I do think that what we can do now with the mobile edge is much different from what we could do just a few years ago. The options for customers are opening up. Customers are moving quickly to leverage both network and application assets much more aggressively. For instance, now they can optimize Wi-Fi to create fast lanes. And they can optimize the applications themselves so they can build a more localized computing experience. And now they can do this, in some cases, for less money and with more control and more security. They can use analytics more precisely to create higher-performing localized environments for the business, versus a larger, centralized environment where it?s very hard to make massive changes without a lot of disruption. It makes companies more nimble. It creates agility in the delivery of applications. And it can provide, in some cases, a more secure environment. Monica: In terms of security, do enterprises feel comfortable about moving more of their functionality to their premises? Matt: I see security as above the distinction between a distributed network with edge computing, and a centralized network. I look at security as its own silo, its own platform. The answer is, yes, some organizations feel that, if they keep things more local, they can control the security component more easily. That premise, in my mind, isn?t inaccurate, but it is problematic. The security framework needs to be rich and robust whether your network is centralized or everything?s at the edge. At Verizon, we believe that our service needs to be highly secure in all cases. Some customers believe that, along with the performance benefits, mobile edge computing offers that high level of security. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |86| Monica: What kind of applications does the enterprise usually need to put closer to the edge? Matt: The internet of things is one driver. Data itself is also driving the move to edge computing. There?s so much data generated from network analytics these days that we can?t physically get all of it across the network. By doing some of the computing at the edge, we lower the latency. Mobile edge computing offers the lower latency that we need to provide the best experience. The more mission-critical an application and the more crucial uptime reliability is to the customer, the more benefit edge computing offers. Edge computing lets us create fast lanes in a highly-optimized Wi-Fi environment and further optimize applications within those fast lanes. We can ensure that the users get a higher quality of service in that environment than they would if we did everything centralized over a giant MPLS network. Monica: Does that include only data, or voice as well? Matt: It?s voice as well. Voice is now a consideration for many businesses that are looking to move everything onto IP and deliver it on whatever form factor fits the business need. It could be a smartphone with an integrated dialer that manages your desk phone and your mobile phone. It could be a wireline phone that is integrated with your mobile number. It could be a VaaS, or video as a service, for services such as video conferences. Edge computing could help deliver with less latency. Monica: New business models may emerge as services become location aware or location based. Do you think the enterprise is willing to pay for at least some of the capex or opex required to deploy local services? Matt: The model?s different, you?re absolutely right. Edge computing creates a capex model. Companies have been moving to more of an opex, computing-as-a-service model. Some customers would prefer to move out of an opex model to a capex model because there are tax implications. And edge computing is a more capital-intensive model. But it?s also important to note that companies like Cisco, as an example, have already built into their operating system, from a networking and routing perspective, the ability to do mobile edge computing and create Wi-Fi fast lanes, and the ability to tag and ensure certain applications have a higher quality of service. Organizations that have invested in this sort of product can pivot and turn on a stronger edge computing experience without a giant checkbook effort. An example would be on tablets in an edge environment. Apple tablets now come with the ability to do application optimization. They come with the ability, using Wi-Fi, to provide a higher quality of service. As the industry starts to look at the benefits of edge computing versus centralized computing as a service, we?re finding that some of the capabilities needed to do it are already in place. Monica: The ability to integrate Wi-Fi with cellular also should be a priority for the enterprise. Matt: It?s very important that the experience inside the four walls is replicated outside the four walls, at least for the critical applications and the optimization needed to make that happen. And we can do that by keeping applications local. But at the same time, Verizon believes that when you leave those four walls, we need to provide the same level of service for that end user that they get within that environment. We spend a good deal of time working with Cisco and all of our networking partners, to ensure that if you migrate onto a 4G LTE connection using 4G LTE Advanced, we can provide a similar look and feel of the optimized model you enjoy now across wireless interfaces, and do it locally. We try to make that experience seamless for the end user, especially when we integrate our VoIP and video services. If they?re part of the mobile edge computing experience, we want to replicate that as they move outside of the four walls into the more centralized approach in the wide-area network. These are the capabilities that we provide customers today. Monica: Again, it?s crucial to have the seamless connectivity because, as an end user, you shouldn?t need to know what access you are using. Edge computing allows you to be access- technology neutral, in the sense that you focus on the functionality, and then it doesn?t matter what RAT the subscriber uses. Matt: We have a strong opinion on why this environment is better with Verizon. The last thing we want is for our end customers to have an experience that?s different. They may understand they?re on cellular, and not on Wi-Fi, but we do not REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |87| want them to notice the difference. And we feel Verizon is the best at doing this. Our customers really don?t know that they?ve moved into an environment in which their services are being provided to them by the Google or Amazon Cloud versus being serviced locally. We work hard with our OEM partners to create a networking environment in which you can move in and out and still get access with the same quality and level of service that you get from a mobile edge environment. Monica: Can you say something about IoT? There?s a huge amount of interest, maybe some hype as well. What do you hear from the enterprise? Matt: IoT is driving companies to edge computing, because of the analytics generated out of an IoT environment, especially in the industrial internet space. Some of our customers are moving more diligently, more pervasively, into sensors and monitoring, and using that data to make better decisions. It?s problematic to transmit the data across traditional networks. Mobile edge computing becomes almost a requirement because we have to keep the data local. We can mine the data and only move the high-level analytics data across a centralized approach. We find that IoT is driving a mobile edge computing experience for data collecting, for the processing of data, especially in certain industries. Think about what GE?s doing, adding sensors to machines. They?re making what I call dumb machines into smart machines so they can gather the data and make better decisions. This is driving mobile edge computing. This level of need has been one of the bigger catalysts for edge computing. Monica: Are there any applications or any specific verticals that are ahead in the move to IoT? Matt: We expect the manufacturing vertical to pick up quickly. We?re seeing some in the utilities vertical, which would include energy. Next would be transportation ? not just transportation and shipping, but also receiving. It?s moving in and out of those four walls, tracking and monitoring all inventory, and reporting in near real-time. In healthcare too, where we can help monitor medicine shipments that need to remain below a certain temperature, and where highly restricted pharmaceuticals like OxyContin need to be constantly managed to help avoid theft or counterfeit. These applications are driving mobile edge computing because of the analytics generated and its use in near real-time decision-making. Monica: You?ve talked a lot about analytics. That?s an interesting part because, oftentimes, we think about edge computing as more secure and residing more on the content side, but also it allows you to manage your network resources better, to optimize them better. Do you think a lot of the analytics is also moving to the edge because, as you say, it?s much more efficient? You?re trying to optimize so you have all the data there. No point sending it all the way back to the core. Matt: Our customers have not completely run to an edge computing mode. It?s a hybrid approach. They still have access to cloud assets that are not necessarily in a centralized architecture. These aren?t going away. What?s going away from the centralized cloud is the massive amount of data being generated locally. We are consuming data locally and evaluating it locally for decision-making. We?re transmitting much smaller data sets up into the cloud. The cloud environment hasn?t dried up or disappeared. It?s still there, but we?re leveraging edge computing so we don?t have to move all of that data up into the cloud to make it happen. Monica: Has it made it cost effective for Verizon, as well, because you?re going to have less cost for transport and backhaul? Matt: Right. The telecommunications capabilities here are remarkable. The ability to create data sets and analyze the data, it?s even outpaced that. What I?m suggesting is that IoT, especially, but mobile computing in general has created such a large set of data that it has outpaced our ability to move it all into a cloud. But it hasn?t outpaced our ability to secure it over the network during transport. Analyzing it locally and then moving what?s needed to the cloud is really the process that we see starting to happen. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |88| Monica: One final question: As you look at the next five years, what do you think will change in mobile edge computing? What new challenges are you going to try to address? Matt: My crystal ball tells me that you?re starting to see network technologies that may swing the pendulum back to a more centralized approach. What?s not going to stop is IoT and the ability to mine data for insights into what it can offer organizations. This is not stopping, it?s accelerating. The volume of data is growing and all of this will only continue exponentially. Also, standards are finally coming into place, both for mobile edge computing and for 5G. As our backhaul is much stronger with fiber, we?re connecting smart cities and smart businesses with much higher-performance networks. And we?ll be using 5G, which has very low latency with incredible speeds. As a result, we open up the ability to move more of that data back to a centralized location. I see the idea of a hybrid IT or an organizational model that meets exactly what each business needs. If you?re a movie studio, you?re going to do more edge computing because you?re going to deal with video that sits within the studio itself. You?re not going to move it, because you need to take action on it there. If you?re a manufacturer, you may need a large capital investment to deploy all the edge computing you need to leverage some of the generalized IT environments. Because of these points, with 5G I have the pipes that are secure enough and have the low-latency capability to move data to whatever environment I need it in and to service those applications. These are disruptive technologies. 5G will disrupt mobile edge computing. Mobile edge won?t go away, but its trajectory might be disrupted because of the new things we can do as we move forward. REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |89| About Verizon Wireless Verizon helps organizations achieve better business outcomes and drive better customer experiences, simply, securely and reliably. With our investments in superior technology like LTE advanced, America?s largest and fastest 4G LTE ever, we deliver innovative solutions like mobility, IoT, cloud, security and telematics that can help you connect people, places and things around the world. About Matt Montgomery Matt Montgomery is the director of marketing for Verizon?s Wireless Business Group. bout BT REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |90| Glossary 3GPP Third Generation Partnership Project ABR Adaptive bitrate [streaming] API Application programming interface ATCA Advanced Telecommunications Computing Architecture AWS Amazon Web Services BBU Baseband unit B-RAS Broadband remote access server BTS Base transceiver station CBRS Citizens Broadband Radio Service CDN Content delivery network CMTS Cable modem termination system CORD Central Office Re-Architected as a Datacenter COTS Commercial off-the-shelf [hardware] CPE Customer premises equipment CPRI Common public radio interface CPU Central processing unit C-RAN Cloud RAN DAS Distributed antenna system DDoS Distributed denial of service DNS Domain name system DPI Deep packet inspection eNodeB Evolved NodeB EPC Evolved Packet Core Eth Ethernet ETSI European Telecommunications Standards Institute FDD Frequency division duplex FLIPS Flexible IP-based Services FPGA Field-programmable gate array GGSN Gateway GPRS support node GPRS General Packet Radio Service GPU Graphics processing unit GTP GPRS Tunneling Protocol GW Gateway HD High definition HeNB Home eNB HSS Home subscriber server HTTPS Hypertext Transfer Protocol Secure I/O Input/output ICN Information-centric networking IMT International Mobile Telecommunications IoT Internet of things IP Internet Protocol IPsec Internet Protocol security IT Information technology L1 [OSI] layer 1 L2 [OSI] layer 2 L3 [OSI] layer 3 LTE Long Term Evolution ME Mobile edge MEC Multiple-access Edge Computing MIMO Multiple input, multiple output MME Mobility management entity MVNO Mobile virtual network operator NAP Network access point NEBS Network Equipment-Building System NFV Network Functions Virtualization NFVI NFV infrastructure NGMN Next Generation Mobile Networks [Alliance] NR New radio NSF National Science Foundation OAM Operations, administration and maintenance OCP Open Compute Project OCR Optical character recognition OEM Original equipment manufacturer OTT Over the top OVP Open Virtualization Profile PAWR Platforms for Advanced Wireless Research PGW Packet gateway PoC Proof of concept PoE Power over Ethernet PTN Public telephone network QoE Quality of experience RAM Random access memory RAN Radio access network RAT Radio access technology RAU Radio aggregation unit RNC Radio network controller RNIS Radio network information service ROI Return on investment RRH Remote radio head REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |91| SDK Software development kit SDN Software-defined networking SD-WAN Software-defined wide area network SGW Serving gateway SP Service Provider STB Set-top box TCO Total cost of operation TCP Transmission Control Protocol TDD Time-division duplex TG Throughput guidance TGE TG entity TIP Telecom Infra Project TLS Transport Layer Security UE User equipment UHD Ultra-high-definition television URLLC Ultra-reliable low-latency communications vBRAS Virtual BRAS vCPE Virtual CPE vEPC Virtual Evolved Packet Core VM Virtual machine VNF Virtualized network function VoD Video on demand VoLTE Voice over LTE VPN Virtual Private Network vRAN Virtual RAN REPORT Power at the edge ?2017 Senza Fili Consulting ? www.senzafiliconsulting.com |92| References [1] 5G Americas, Understanding information centric networking and Mobile Edge Computing, 2016. [2] 5G Public-Private-Partnership, The 5G Infrastructure Public Private Partnership: The next generation of communication networks and services, 2015. [3] 5G Vision ? The 5G Infrastructure Public Private Partnership: The next generation of communication networks and services, Release 5.1, 5G Public-Private-Partnership, 2015. [4] Beck, Michael Till, Martin Werner, Sebastian Feld, and Thomas Schimper, Mobile Edge Computing: A taxonomy, AFIN 2014. [5] Bhardwaj, Ketan, Ming-Wei Shih, Pragya Agarwal, Ada Gavrilovska, Taesoo Kim, and Karsten Schwan, Fast, scalable and secure onloading of edge functions using AirBox, Georgia Institute of Technology, 2016. [6] Chao Hu, Yun, Milan Patel, Dario Sabella, Nurit Sprecher, and Valerie Young, Mobile Edge Computing: A key technology towards 5G, White paper 11, ETSI, 2015. [7] Cisco, Cisco Visual Networking Index: Global Mobile Data. Traffic Forecast Update, 2016?2021, 2017. [8] Ericsson, Ericsson Mobility Report, 2016 [9] ETSI, Mobile Edge Computing (MEC) Terminology, ETSI GS MEC 001 V1.1.1, 2016. [10] ETSI, Mobile Edge Computing (MEC): Technical requirements, ETSI GS MEC 002 V1.1.1, 2016. [11] ETSI, Mobile-Edge Computing (MEC): Service scenarios, ETSI GS MEC-IEG 004 V1.1.1, 2015. [12] ETSI, Mobile-Edge Computing (MEC): Proof of concept framework, ETSI GS MEC-IEG 005 V1.1.1, 2015. [13] ETSI, Mobile Edge Computing: Market acceleration: MEC metrics best practice and guidelines, ETSI GS MEC-IEG 006 V1.1.1, 2017. [14] ETSI, Mobile-Edge Computing, 2014. [15] Friedman, Thomas L., Thank You for Being Late: An Optimist's Guide to Thriving in the Age of Accelerations, 2016 [16] I, Chih-Lin, Corbett Rowell, Shuangfeng Han, Zhikun Xu, Gang Li, and Zhengang Pan, Toward green and soft: A 5G perspective, IEEE Communications Magazine, 2014. [17] I, Chih-Lin, Shuangfeng Han, Zhikun Xu, Qi Sun, and Zhengang Pan, 5G: Rethink mobile communications for 2020+, Philosophical Transactions of the Royal Society A, 2016. [18] Intel, Real-world impact of Mobile Edge Computing (MEC), 2016. [19] InterDigital, What the MEC? An architecture for 5G. [20] International Telecommunication Union, IMT vision ? Framework and overall objectives of the future development of IMT for 2020 and beyond, Recommendation ITU-R M.2083-0, 2015. [21] International Telecommunication Union, IMT vision: Framework and overall objectives of the future development of IMT for 2020 and beyond, Recommendation ITU-R M.2083-0, 2015. REPORT Massively densified networks ? 2016 Senza Fili Consulting ? www.senzafiliconsulting.com |93| [22] Marek, Peter, Delivering elasticity to the mobile edge, Embedded Innovator, 2016. [23] Miller, Linsey, How can operators make money with MEC? Artesyn.com Newsroom, 2016. [24] Paolini, Monica, Charting the path to RAN virtualization: C-RAN, fronthaul and HetNets, Senza Fili, 2015. [25] Paolini, Monica, Massively densified networks: Why we need them and how we can build them, Senza Fili, 2016. [26] Paolini, Monica, The smart RAN: Trends in the optimization of spectrum and network resource utilization, Senza Fili, 2015. [27] Qwilt, Open Edge Cloud business case, 2016. [28] Reister, John, It?s MEC to the rescue for struggling mobile video performance, Telecoms.com, 2017 [29] Vasona Networks, Are mobile networks ready for the next gaming craze? Vasona Networks, 2016 REPORT Massively densified networks ? 2016 Senza Fili Consulting ? www.senzafiliconsulting.com |94| Further resources Latest reports in this series: Improving latency and capacity in transport for C-RAN and 5G. Trends in backhaul, fronthaul, xhaul and mmW Massively densified networks. Why we need them and how we can build them Voice comes to the fore, again. VoLTE and Wi-Fi Calling redefine voice Getting the best QoE: Trends in traffic management and mobile core optimization The smart RAN. Trends in the optimization of spectrum and network resource utilization Charting the path to RAN virtualization: C-RAN, fronthaul and HetNets LTE unlicensed and Wi-Fi: moving beyond coexistence Watch the video of the interviews About RCR Wireless News Since 1982, RCR Wireless News has been providing wireless and mobile industry news, insights, and analysis to industry and enterprise professionals, decision makers, policy makers, analysts and investors. Our mission is to connect, globally and locally, mobile technology professionals and companies online, in person, in print and now on video. Our dedication to editorial excellence coupled with one of the industry?s most comprehensive industry databases and digital networks leads readers and advertisers to consistently choose RCR Wireless News over other industry publications. About Senza Fili Senza Fili provides advisory support on wireless data technologies and services. At Senza Fili we have in-depth expertise in financial modelling, market forecasts and research, white paper preparation, business plan support, RFP preparation and management, due diligence, and training. Our client base is international and spans the entire value chain: clients include wireline, fixed wireless and mobile operators, enterprises and other vertical players, vendors, system integrators, investors, regulators, and industry associations. We provide a bridge between technologies and services, helping our clients assess established and emerging technologies, leverage these technologies to support new or existing services, and build solid, profitable business models. Independent advice, a strong quantitative orientation, and an international perspective are the hallmarks of our work. For additional information, visit www.senzafiliconsulting.com or contact us at info@senzafiliconsulting.com or +1 425 657 4991. About the author Monica Paolini, PhD, is the founder and president of Senza Fili. She is an expert in wireless technologies and has helped clients worldwide to understand new technologies and customer requirements, create and assess financial TCO and ROI models, evaluate business plan opportunities, market their services and products, and estimate the market size and revenue opportunity of new and established wireless technologies. She frequently gives presentations at conferences, and writes reports, blog entries and articles on wireless technologies and services, covering end-to-end mobile networks, the operator, enterprise and IoT markets. She has a PhD in cognitive science from the University of California, San Diego (US), an MBA from the University of Oxford (UK), and a BA/MA in philosophy from the University of Bologna (Italy). You can reach her at monica.paolini@senzafiliconsulting.com. ? 2017 Senza Fili Consulting, LLC. All rights reserved. The views and statements expressed in this document are those of Senza Fili Consulting LLC, and they should not be inferred to reflect the position of the report sponsors, or other parties participating in the interviews. No selection of this material can be copied, photocopied, duplicated in any form or by any means, or redistributed without express written permission from Senza Fili Consulting. While the report is based upon information that we consider accurate and reliable, Senza Fili Consulting makes no warranty, express or implied, as to the accuracy of the information in this document. Senza Fili Consulting assumes no liability for any damage or loss arising from reliance on this information. Names of companies and products here mentioned may be the trademarks of their respective owners. Cover photo by Senza Fili, Loyly, Helsinki.