Monday, May 25, 2020

Why telco operators need a platform for edge computing

Initially published in The Mobile Network.

Extracted from the edge computing and hybrid cloud 2020 report.

Edge computing and hybrid clouds have become subjects of many announcements and acquisitions over the last months.
Edge computing, in order to provide a capacity for developers and third party to reserve and consume operators computing, storage and networking capacity need a platform. The object of this platform is to provide a web interface and series of APIs to abstract network topology and complexity and offer developers a series of cloud services and product to package within their offering. Beyond hyperscalers who have natively developed these platforms, a few vendors have emerged in the telco space, such as MobiledgeX and ORI Industries.
Network operators worldwide are confronted with the inexorable growth of their data traffic due to the consumers’ voracious appetite for video streaming and gaming. Since video content is the largest and fastest growing data type in the networks, an economical challenge is slowly arising. Data charging models have departed from per Megabyte metered billing to bundles and unlimited data, which encourages traffic growth, while reducing the operators’ capacity to monetize this growth. Consumers are not willing to pay much more for a HD video versus Standard Definition. For them, it is essentially the same service and the operator is to blame if the quality is not sufficient. Unfortunately, the problem is likely to accelerate with emerging media hungry video services relying on 4K, 8K and Augmented Reality. As a consequence, the average revenue per user stagnates in most mature markets, while costs continue to rise to increase networks capacity.
While 5G promises extraordinary data speeds, enough to complement or equal fibre fixed capacity, there is no real evidence that the retail consumer market will be willing to pay a premium for improved connectivity. If 5G goes the way of 4G, the social media, video streaming, gaming services and internet giants will be the ones profiting from the growth in digital services. The costs for deploying 5G networks will range in the low to double digit billions, depending on the market, so… who will foot the bill?
If properly executed, the 5G roll out will become in many markets the main broadband access at scale. As this transition occurs, new opportunities arise to bundle mobile connectivity with higher level services, but because the consumer market is unlikely to drastically change its connectivity needs in the short term, the enterprise market is the most likely growth opportunity for 5G in the short to medium term.
Enterprise themselves are undergoing a transformation, with the commoditization of cloud offering.
Cloud is one of the fastest growing ICT businesses worldwide, with IaaS the fastest growing segment. Most technology companies are running their business on cloud technology, be it private or public and many traditional verticals are now considering the transition.
Telecom operators have mostly lost the cloud battle - AWS, Microsoft, Google, Alibaba have been able to convert their global network of data centers into an elastic, on-demand as-a-service economy.
Edge computing, the deployment of mini data centers in telco networks promises to deliver a range of exciting new digital services. It may power remote surgery, self driving cars, autonomous industrial robots, drone swarms and countless futuristic applications.
In the short term, though, the real opportunity is for network operators to rejoin the cloud value chain, by providing a hyper local, secure, high performance, low latency edge cloud that will complement the public and private clouds deployed today.
Most private and public clouds ultimately stumble upon the “last mile” issue. Not managing the connectivity between the CPE, the on-premise data center and the remote data center means more latency, less control and more possibility for hacking or privacy issues.
Operators have a chance to partner with the developers’ community and provide them with a cloud flavour that extends and improve current public and private cloud capabilities.
The edge computing market is still emerging, with many different options in terms of location, distribution, infrastructure and management, but what is certain is that it will need to be more of a cloud network than a telco network if it succeeds in attracting developers.
Beyond the technical details that are being clarified by deployments and standards, the most important gap network operators need to bridge with a true cloud experience is the platform. Operators traditionally have deployed private cloud for their own purpose -  to manage their network. These clouds do not have all the traditional features we can expect from commercial public cloud (lifecycle management, third party authentication, reservation, fulfillment…). The key for network operators to capture the enterprise opportunity is to offer a set of APIs that are as simple as those from the public clouds, so that developers and enterprise may reserve, consume and pay for edge computing and connectivity workloads and pipelines.
A possible outcome of this need if operators do not open their private cloud to enterprises is that hyperscalers will expand their clouds to operators’ networks and provide these services to their developer and client community. This would mean that operators would be confined to a strict connectivity utility model, where traffic prices would inexorably decline due to competitive pressure and high margin services would be captured by the public cloud.
  • Edge computing can allow operators to offer IaaS and PaaS services to enterprises and developers with unparalleled performance compared to traditional clouds:
  • Ultra-low and guaranteed latency (typically between 3 -25ms between the CPE and the first virtual machine in the local cloud)
  • Guaranteed performance (up to 1Gps in fibre and 300Mbps in cellular)
  • Access to mobile edge computing (precise user location, authentication, payment, postpaid / prepaid, demographics… depending on operators’ available APIs)
  • Better than cloud, better than WIFI services and connectivity (storage, video production, remote desktop, collaboration, autonomous robots,…)
  • Flexible deployment and operating models (dedicated, multi-tenant…)
  • Local guaranteed data residency (legal, regulatory, privacy compliant)
  • Reduce cloud costs (data thinning and preprocessing before transfer to the cloud)
  • High performance ML and AI inferring
  • Real time guiding and configuration of autonomous systems

It is likely that many enterprise segments will want to benefit from this high-performance cloud. It is also unlikely that operators alone will be able to design products and services for every vertical and segment. Operators will probably focus on a few specific accounts and verticals, and cloud integration providers will rush in to enable specific market edge cloud and connectivity services:
  • Automotive
  • Transport
  • Manufacturing
  • Logistics
  • Retail
  • Banking and insurances
  • IoT
  • M2M…

Each of these already have connectivity value chain, where network operators are merely a utility provider for higher value services and products. Hybrid local cloud computing offer the operators the opportunity to go up the value chain by providing new and enhanced connectivity and computing products directly to consumers (B2C), enterprises (B2B) and developers (B2B2x).

Fixed and mobile networks have not been designed to expose their capabilities to third party for reservation, consumption and payment of discrete computing and connectivity services. Edge computing, as a new greenfield environment is a great place to start if an operator would like to offer these types of services. Because it is new, there is no legacy deployed and the underlying technology is closer to cloud native. This is necessary to create a developer and enterprise platform. Nonetheless, an abstraction layer is necessary to federate and orchestrate the edge compute infrastructure and provide a web-based authentication, management, reservation, fulfillment, consumption and payment model for enterprises and developers to contract these new telco services.
This is what a platform provides. An abstraction layer, that hides telco networks complexity, federates all edge computing capacity across various networks and operators and present a coherent marketplace for enterprise and developers to build and consume new services offered by the operator community as IaaS, PaaS and SaaS. By deploying a platform, operators can reintegrate the cloud supply chain, but they will have to decide whether they want to own the developer relationship (and build their own platform) or benefit from existing ecosystems (and deploy an existing third party platform). In the first case, it is a great effort, but the revenues flow directly to the operator, the platform is just another technology layer. In the second, revenues go to the platform provider and are shared with the operator. It provides faster time to market, but less control and margin. This model, in my mind is inevitable, it remains to be seen whether operators will be able to develop and deploy the first one in time and at scale.

Monday, May 11, 2020

Why Telcos need Open Core Surgery

 (This article was initially published in Light Reading)

At Mobile World Congress, TIP (the Telecom Infra Project, an industry forum created by Facebook and a number of leading telco operators and IT vendors), announced the creation of a new project group called Open Core Network. Details have starting to emerge last week, with a webinar.
The ambitious target of the group is to define and develop an open and disaggregated 4G Evolved Packet Core and 5G Core for wireless, wired, Wi-Fi on a variety of use cases.

We have seen in the recent past that various attempts to open up the telco cloud ecosystem and value chain have had contrasted results.
  • Telco clouds, based on VNFs and Openstack-like virtualization layer have mostly failed to reach critical mass in deployment and usability.
  •  ETSI-defined orchestration efforts based on open source projects such as OSM (Open Source Mano) and ONAP (Open Network Automation Platform) have been a work in progress and have equally, to date, failed to become automated telco networks app stores.
  • TIP has been successful with the definition, launch and deployment of Open RAN. We have recently seen announcements from Altiostar, Nokia and Cisco in Rakuten's network, as well as from Mavenir in Idea and DISH networks.

As we know, these efforts are aimed at disrupting the current telecom infrastructure provider cost structure by disaggregating traditional networks.
First by separating hardware from software, so that the solutions can be deployed in white boxes - Commercial Off The Shelf (COTS) hardware - rather than costly proprietary ones.
Second by breaking telecom functions into software elements that can be deployed, managed and sourced independently from each other. This is key in the sense that it allows new vendors to enter the ecosystem, who can specialize in specific elements rather than end-to-end solutions. This increases competition and allow a more flexible sourcing strategy, with either best-of-breed vendors for each elements or selection of vendors for fit-for-purpose use cases deployments. The key to enable this scenario is an architecture that is accepted by all, with well-defined software elements functions and more importantly, open, standard, rigid interfaces that guarantee that one vendor can be substituted by another without undue integration effort.

5G is supposed to be the first telco cloud network that is natively virtualized, software-defined, elastic and automated at scale. This can be achieved today by deploying a single vendor solution from one of the dominant telco vendors. Things start to complicate vastly if one wants to deploy a multi-vendor network. Since the standards are not quite finalized on some of the elements and behaviour of a 5G network and operators are announcing and launching 5G networks nonetheless, vendors have to fill the gaps with proprietary implementations, and extensions to the standards to make their end-to-end solution automated, software defined and elastic.

One last bastion of telco proprietary implementation is the Core network. The Core network is basically the brain of the telco network. All the consumer data is stored there, all the charging systems reside there, all the elements to decide where traffic should go and how it should be treated live in the Core. This brain is very complex and composed of a number of elements that have, until now, usually been sold and deployed from single vendors. This has long been a trojan horse for dominant telco vendors to control a network. It is also a self-perpetuating decision, as the evolution from one standard version to another or from one generation to another is much more cost effective as an upgrade of the current vendor's solution as opposed to a rip and replace by a new vendor. 
With 5G, the traditional vendors had a few different architectural options for Core deployment and they mostly elected a non-standalone (NSA) version, which can only be deployed as an upgrade to the 4G EPC. It essentially guarantees that a current 4G Core deployment will evolve to 5G with the same vendor, perpetuating the control over the network. This does not only affect the Core network, it also affects the Radio Access Network (RAN), as its implementation, in the early stage of 5G is dependent on an harmonious interworking with the Core. As a result, many traditional Core vendors who are also RAN vendors have created a situation where the only practical and economical way for an operator to launch 5G fast is to deploy Core and RAN from that same vendor. This situation perpetuates the oligopoly in telco supply chain, which reduces innovation and increase costs.

TIP's Open Core is an attempt to create a Core network for 4G and 5G that will be open, composed of software elements that will be provided by independent vendors, all using the same open interfaces to allow low-touch integration and increase the rate of innovation. If the group follows the same path as Open RAN, it could become a major disruption in telco networks, enabling for the first time in decades the possible deployment of a full telco network from a rich ecosystem of vendors and an innovation pace in sync with what we have seen from the hyperscaler world.

Friday, May 8, 2020

What are today's options to deploy a telco cloud?

Over the last 7 years, we have seen leading telcos embracing cloud technology as a mean to create an elastic, automated, resilient and cost effective network fabric. There has many different paths and options from a technological, cultural and commercial perspective.

Typically, there are 4 categories of solutions telcos have been exploring:

  • Open source-based implementation, augmented by internal work
  • Open source-based implementation, augmented by traditional vendor
  • IT / traditional vendor semi proprietary solution
  • Cloud provider solution

The jury is still out as to which option will prevail, as they all have seen growing pains and setbacks.

Here is a quick cheat sheet of some possibilities, based on your priorities:

Obviously, this table changes quite often, based on progress and announcements of the various players, but it can come handy if you want to evaluate, at high level, what are some of the options and pros / cons of deploying one vendor or open source project vs another.

Details, comments are part of my workshops and report on telco edge and hybrid cloud networks.