Thursday, April 23, 2020

Hyperscalers enter telco battlefront

We have, over the last few weeks, seen a flurry a announcements from hyperscalers investing in telco infrastructure and networks. Between Facebook's $5.7B investment in India's Jio Reliance, to Microsoft's acquisition of Affirmed Networks for $1.35B or even AWS' launch of Outpost and Google's Anthos ramp up.


Why are hyperscalers investing in telecom gear and why now?

Facebook had signalled its intent as far as 2016 when Mark Zuckerberg presented at mobile world congress his vision for the future of the company.


Beyond the obvious transition from picture and video sharing to virtual / augmented reality, tucked-in in the top right, are two innocuous words “telco infra”.
What Facebook realized is that basically anyone who has regular access to broadband will likely use a Facebook service. One way to increase the company’s growth is to invent / buy / promote more services, which is costly and uncertain. Another way is simply to connect more people.
With over 2,5 billion Facebook products users, the company still has some space to grow in this area, but the key limiting factor seems to be connectivity itself. The last billions of broadband unconnected are harder to attain because traditional telecom networks do not reach there. The last unconnected are mostly in rural area. Geographically disperse, with a lower income than their urban counterparts.
Looking at this problem from their perspective, Facebook reached a similar conclusion to the network operators operating in these markets. Traditional telco networks are too expensive to deploy and maintain to reach this population sustainably. The same tactics employed by operators to disaggregate and stimulate the infrastructure market can be refocused and better stimulated by Facebook.
This was the start of Facebook Connectivity, a specific line of business in the social media’s giant empire to change the cost structure of telco networks. Facebook connectivity has evolved to encompass a variety of efforts, ranging from the creation of TIP (an open forum to disaggregate and open telco networks), the co investment with Telefonica in a Joint Venture dedicated to connect the unconnected in latin america and this week, the announcement of its acquisition of 9.9% of Jio Reliance in India.


How about Microsoft, Google and others?

Google had, before the recent open source cloud platform Anthos dug their toes in telco water with project Fi and its fiber businesses.
Microsoft has been trying for he last 5 years to exploit the transition in telco networks from proprietary to IT. Even IBM's Redhat acquisition had a telco interest, as the giants also try to become a more prevalent vendor in the telco ecosystem.

So... why now?

Another powerful pivot point in Telecom is the emergence of 5G. As the latest telephony technology generation rolls out, telco networks are undeniably being re-architected and redesigned to look more like cloud networks. This creates an interesting set of risks and opportunities for incumbents and new entrants alike.
For operators, the main interest is to drastically reduce the cost of rolling out and maintaining complex telco networks by using powerful virtualization, SDN and automation techniques that have allowed hyperscalers to dominate cloud computing. These technologies, if applied correctly can transform the cost structure of network operators, particularly important at the outset of multi billion dollars investment in 5G infrastructure. The radical cost structure disruption comes from disaggregation of the network between hardware and software, the introduction of new vendors in the value chain who drive price pressure on incumbents and the widespread automation and cloud economics.
These opportunities bring also new risks. While they open up the supply chain with the introduction of new vendors, they also allow new actors to enter the value chain, either to substitute and dominate legacy vendors or create new control points (see the orchestrator wars I have been mentioning in previous posts). The additional risk is that the cost of entry into telco becomes lower for cloud hyperscalers as the technology to run telco networks transitions from proprietary closed ecosystem to open source, cloud environment.

The last pivot point is another telco technology that is very specifically aimed at creating a cloud environment in telco networks: Edge computing. It creates a cloud layer that can allow the provision, reservation and consumption of telco connectivity, together with cloud computing. As a greenfield environment, it is a natural entry point for cloud operators and new vendors alike to enter the telco ecosystem.

Facebook, Google, AWS, Microsoft and others seem to think that 5G and edge computing in particular will be more cloud than telco. Network operators try to resist this claim by building a 5G network that will be a fully integrated connectivity and computing experience, complementary to public clouds, but different enough to command a premium, a different value chain and operator control.

In which direction will the market move? This and more in my report and workshop Edge computing and Hybrid Clouds 2020.

Wednesday, April 15, 2020

The business cases of edge computing

Edge computing has been a trendy topic over the last year. Between AWS' launch of Outpost, Microsoft continuous effort with Azure Stack, Nvidia's specialized gaming version EGX platform or even Google's Anthos toolkit, much has been said about this market segment.
Network operators, on their side, have announced plans for deployments in many geographies, but with little, in terms of specific new services, revenues or expected savings.
Having been in the middle of several of these discussions, between vendors, hyperscalers, operators and systems integrators, I am glad to share a few thoughts on the subject.

Hyperscalers have not been looking at edge computing as a new business line, but rather as an extension of their current cloud capabilities. There are many use cases today that cannot be fully satisfied by the cloud, due to a combination of high / variable latency, network congestion, and lack of visibility / control of the last mile connectivity.
For instance, anyone having tried to edit online a diagram in powerpoint office 365 or to play a massive multiplayer online cloud game will recognize how maddeningly frustrating the experience can be.
Edge computing, as in bringing cloud resources closer physically to where data is consumed / produced makes sense to reduce latency and the need for on-premise dedicated resources. From an hyperscaler's perspective, edge computing can be as simple as dropping a few racks within an operator data center to allow their clients to use and configure new availability zones with specific performance and price.

Network operators, who have largely lost the cloud computing wholesale market to the hyperscalers, see edge computing as an opportunity to reintegrate the value chain, by offering cloud-like services at incomparable performance. Ideally, they would like to capture and retain the emerging high performance cloud computing market that will be sure to spurn a new category of digital services ranging from AI-augmented manufacturing and automation, autonomous vehicles, ubiquitous facial and object recognition and compute-less smart devices. The problem is that a lot of these hypothetical services are ill-defined, far fetched and futuristic, which does not inspire sufficient confidence to the CFO that has to approve multi - billion capital expenditure to get going.
But surely, if the likes of Microsoft, Intel, HP, Google, Facebook, AWS are investing in Edge Computing there must be something there? What are the operators missing to make the edge computing business case positive?

Mobile or multi access edge computing?

Many operators looked at edge computing first from the perspective of mobile. The mobile edge computing business case remains extremely uncertain. There is no identified use case that justifies the cost to deploy thousands of mini compute capabilities at mobile site in the short term. Even with the perspective of upgrading networks to 5G, the added cost of mobile edge computing is hard to justify.

If not in mobile site, the best bet to deploy edge computing for network operators is in Central Offices (CO). These facilities house commuting platforms for copper, fiber, DSL connectivity and are overdue for upgrade in many markets. The deployment of fibre, the copper replacement and the evolution of technology from GPON to XGS-PON and PON2 are excellent windows of opportunity to replace aging single-purposes infrastructure with open, software defined computing capability.
The level of investment for central offices retooling into mini data centers is orders of magnitude lower than the mobile case, and is completely flexible. It is not necessary to change all central offices, one can proceed by deploying one per state / province / region and increase capillary as business dictates.

What use cases would make edge computing's business case positive for operators in that scenario?


  • First, for operators who have triple and quadruple play, the opportunity to replace aging dedicated infrastructure for TV, fixed telephony, enterprise and residential connectivity by cloud native software defined open architecture provides interesting savings and benefits. The savings are realized from the separation of hardware and software, the sourcing and deployment of white boxes and the opex savings of separating control plane and centralizing and automating service elasticity. 
  • Additional savings are to be had with the deployment at the edges of content / video caches. Particularly for TV providers who see the increase of on-demand and unicast live traffic, positioning edge caches allow up to 80% savings in content transport. This is likely to increase with the upgrade from HD to 4K, 8K and growth in AR/VR.
  • At last, for operators who are deploying their CPE in their customers' home, edge computing allows to simplify and reduce drastically the cost of these equipments and their deployment / maintenance by bringing the services into the Central Office and reducing the need for storage and compute in the CPE.

While the savings can be significant in the long run, no operator can justify substituting existing infrastructure if its amortization is not fully realized on these premises alone. This is why some operators are looking at these scenarios only for greenfield fiber deployments or as part of massive copper replacement windows.
Savings alone in all likeliness won't allow operators to deploy at the rhythm necessary to counter hyperscalers. New revenues streams can also be captured with the deployment of edge computing.

  • For consumers, it is likely that the lowest hanging fruit in the short term is in gaming. While hyperscalers and gaming companies have launched their own cloud gaming services, their success has been limited due to the poor online experience. The most successful game franchises are Massive Multiplayer Online. They pitch dozens of players against each other and require a very controlled latency between all players for a fair and enjoyable gameplay. Only operators can provide controlled latency if they deploy gaming servers at the edge. Without a full blown gaming service, providing game caching at the edge can drastically reduce the download time for games, updates and patches, which increases dramatically player's service satisfaction.
  • For enterprise users, edge computing has dozens of use cases that can be implemented today that are proven to provide superior experience compared to the cloud. These services range from high performance cloud storage, to remote desktop, to video surveillance and recognition.
  • Beyond operators-owned services, the largest opportunity is certainly the enablement of edge as a service (EaaS), allowing cloud developers to use edge resources as specific cloud availability zones.
The main issue at this stage, for operators is to decide whether to let hyperscalers deploy their infrastructure in their network, capturing most of the value of these emerging services but also opening up a new line of revenue from wholesale hosting or trying to play it alone, as an operator or a federation of them, deploying a telco cloud infrastructure and building the necessary platform to resell edge compute resource in their networks.

This and a lot more use cases and business cases in my online workshop and report Edge Computing 2020.