Infrastructure Edge: Awaiting Development

Editor’s Note: This article originally appeared on the State of the Edge blog. State of the Edge is a collaborative research and educational organization focused on edge computing. They are creators of the State of the Edge report (download for free) and the Open Glossary of Edge Computing (now an official project of The Linux Foundation).

When I began looking into edge computing just over 24 months ago, weeks would go by with hardly a whimper on the topic, apart from sporadic briefs about local on-premises deployments. Back then, there was no State of the Edge Report and certainly no Open Glossary of Edge Computing. Today, an hour barely passes before my RSS feed buzzes with the “next big announcement” around edge. Edge computing has clearly arrived. When Gartner releases their 2018 Gartner Hype Cycle later this month, I expect edge computing to be at the steepest point in the hype cycle.

Coming from a mobile operator heritage, I have developed a unique perspective on edge computing, and would like to double-click on one particular aspect of this phenomenon, the infrastructure edge and its implication for the broader ecosystem.

The Centralized Data Center and the Wireless Edge

 

So many of the today’s discussions about edge computing ascribe magical qualities to the cloud, suggesting that it’s amorphous, ubiquitous and everywhere. But this is a misnomer. Ninety percent of what we think of as cloud is concentrated in a small handful of centralized data centers, often thousands of miles and dozens of network hops away. When experts talk about connecting edge devices to the cloud, it’s common to oversimplify and emphasize the two endpoints: the device edge and the centralized data center, skipping over the critical infrastructure that connects these two extremes—namely, the cell towers, RF radios, routers, interconnection points, network hops, fiber backbones, and other critical communications systems that liaise between edge devices and the central cloud.

In the wireless world, this is not a single point; rather, it is distributed among the cell towers, DAS hubs, central offices and fiber routes that make up the infrastructure side of the last mile. This is the wireless edge, with assets currently owned and/or operated by network operators and, in some cases, tower companies.

The Edge Computing Land Grab

 

The wireless edge will play a profound and essential role in connecting devices to the cloud. Let me use an analogy of a coastline to illustrate my point.

Imagine a coastline stretching from the ocean to the hills. The intertidal zone, where the waves lap upon the shore, is like the device edge, full of exciting activities and a robust ecosystem, but too ephemeral and subject to change for building a permanent structure. Many large players, including Microsoft, Google, Amazon, and Apple, are vying to win this prized spot closest at the water’s edge (and the end-user) with on-premises gateways and devices. This is the domain of AWS Greengrass and Microsoft IoT Edge. It’s also the battleground for consumers, with products like Alexa, Android, and iOS devices, In this area of the beach, the battle is primarily between the internet giants.

On the other side of the coastline, opposite the water, you have the ridgeline and cliffs, from where you have an eagle view of the entire surroundings. This “inland” side of the coastline is the domain of regional data centers, such as those owned by Equinix and Digital Realty. These data centers provide an important aggregation point for connecting back to the centralized cloud and, in fact, most of the major cloud providers have equipment in these co-location facilities.

And in the middle — yes, on the beach itself — lies the infrastructure edge, possibly the ideal location for a beachfront property. This space is ripe for development. It has never been extensively monetized, yet one would be foolhardy to believe that it has no value.

In the past, the wireless operators who caretake this premier beachfront space haven’t been successful in building platforms that developers want to use. Developers have always desired global reach along with a unified, developer-friendly experience, both of which are offered by the large cloud providers. Operators, in contrast, have largely failed on both fronts—they are primarily national, maybe regional, but not global, and their area of expertise is in complex architectures rather than ease of use.

This does not imply that the operator is sitting idle here. On the contrary, every major wireless operator is actively re-engineering their networks to roll out Network Function Virtualization (NFV) and Software Defined Networking (SDN), along the path to 5G. These software-driven network enhancements will demand large amounts of compute capacity at the edge, which will often mean micro data centers at the base of cell towers or in local antenna hubs. However, these are primarily inward-looking use cases, driven more from a cost optimization standpoint rather than revenue generating one. In our beach example, it is more akin to building a hotel call center on a beachfront rather than open it up primarily to guests. It may satisfy your internal needs, but does not generate top line growth.

Developing the Beachfront

 

Operators are not oblivious to the opportunities that may emerge from integrating edge computing into their network; however, there is a great lack of clarity about how to go about doing this. While powerful standards are emerging from the telco world, Multi-Access Edge Computing (MEC) being one of the most notable, which provides API access to the RAN, there is still no obvious mechanism for stitching these together into a global platform; one that offers a developer-centric user experience.

All is not lost for the operator;, there are a few firms such as Vapor IO and MobiledgeX that have close ties to the  infrastructure and operator communities, and are tackling the problems of deploying shared compute infrastructure and building a global platform for developers, respectively. Success is predicated on operators joining forces, rather than going it alone or adopting divergent and non-compatible approaches.

In the end, just like a developed shoreline caters to the needs of visitors and vacationers, every part of the edge ecosystem will rightly focus on attracting today’s developer with tools and amenities that provide universal reach and ease-of-use. Operators have a lot to lose by not making the right bets on programmable infrastructure at the edge that developers clamor to use.  Hesitate and they may very well find themselves eroded and sidelined by other players, including the major cloud providers, in what is looking to be one of the more exciting evolution to come out of the cloud and edge computing space.

The Un-network Uncarrier

Note: The article is the first of a 3-part series which talks Facebook’s Telecom Infrastructure Partnership (TIP) initiative and the underlying motivation behind the same. The next two parts are focused on the options and approaches that Operators and vendors respectively should adopt in order not to suffer the same outcome as what happened in OCP.

When T-Mobile USA announced its latest results on July 19, 2017, CEO John Legere remarked with delight – “we have spent the past 4.5 years breaking industry rules and dashing the hopes and dreams of our competitors…”. The statement rung out true – in 2013 the stock hovered at a paltry USD 17.5 and now was close to USD 70 – with 17 straight quarters of acquiring more than a 1m customers per quarter. If anything – the “Un-carrier” has yet proven unstoppable. While this track record is remarkable, my hypothesis is that in the long term the “uncarrier” impact may seem puny compared to “un-network” impact which could potentially upend the entire Telecom industry – operators and vendors alike. If you are still wondering what I am referring to – then it is TIP, the open Telecom Infrastructure Initiative started by Facebook in 2016 which now counts more than 300 members.

The motivation for TIP came on the heels of the wildly successful Open Compute Project (OCP). OCP, whose members include Apple, Google and Microsoft heralded a new way to “white-box” a data center based upon freely available specifications and a set of contract vendors ready to implement based upon preset needs. It also heralded a new paradigm (now also seen in other industries) – you didn’t need to build and own it… to disrupt it (think Uber and Airbnb). And while many may falsely be led to believe the notion that this is primarily for developing countries in the light of the Facebook goal to connect the “un-connected”, I believe the impact will be for both – operators in developed and developing countries alike. There are a few underlying reasons here.

• A key ingredient in building a world class product is to have a razor sharp focus towards understanding (and potentially forecasting) your customer/ user needs. While operators do some bit of that, Facebook’s approach raises the bar to an altogether different level. And in this analysis – one thing is obvious (something which operators also are painfully aware of); people have a nearly insatiable quest for data. A majority of this data is in the form of video; unsurprisingly video has become a major focus across all Facebook products – be it Facebook Live, Messenger Video calling or Instagram. The major difference between the developing and developed world in this respect is the expectations of what data rates to expect, and the willingness (in many cases the ability) to pay for the same.
• In the 1^st world nations, 5G is not a pipe-dream anymore and has operators worried about the high costs associated in rolling out and maintaining such a network (not to forget the billions spent in license fees). In order to recoup this, there may be an impulse to increase prices – which could swing many ways; people grudgingly pay; consumers revolt resulting in margin pressure – and lastly – people simply use less data. The last option should worry any player in the video game; what if people elect to skip the video to save costs!
• Among the developing world, 5G may still be someway off, but here in the land of single digit ARPU’s, operators have a limited incentive for heavy investment given a marginal (or potentially negative ROI). On top of that there is a lack of available talent and heavy dependence on vendors – those whose primary revenue source comes from selling even more equipment and services; all increasing the end price to the consumer.

Facebook has it’s fingers in both these (advanced and developing market) pies and while its customers may be advertisers, its strength comes from the network effect created by its billion strong user base and it is essential to provide these with the best experience possible.

These users want great experiences, higher engagement and better interaction. Facebook also knows another relevant data point – that building, owning and operating a network is hard work, asset heavy and not as profitable. It has to look no further than Mountain View where Google has been scaling down its Fiber business. Truck rolls are frankly – not sexy.

It is not that operators were not aware of this issue, but traditionally lacked the knowhow needed to tackle this challenge which required deep hardware expertise. This changed with the arrival of “softwarization” of hardware in the form of network function virtualization. Operators were not software guru’s ….. but Facebook was. It recognized that by using this paradigm shift and leveraging its core software competencies it could potentially transform and potentially disrupt this market.

Operators could leverage the benefit of open source design to vastly drive down the costs of implementing their networks; vendors would no longer have the upper hand and disrupt the current paradigm of bundling software, hardware and services. Implementations could potentially result in a better user experience – with Facebook as one of the biggest beneficiaries. Rather than spend billions in connecting the world, it would support others to do so. In doing so, it would have access to infrastructure that it helped architect without owning any infrastructure. In short – it would be – the “un-network – uncarrier”.