Finding the edge for edge

Phil Marshall / Tolaga Research

The battle-lines are shifting. In the past, the internet and cloud computing pushed value creation away from telecoms networks, and towards end-point devices and cloud data centers. This, to the benefit of web-scale providers and device manufacturers, like Amazon, Apple, Facebook, Google and Samsung, largely at the expense of telecoms network operators. 

Cloud service architectures are centralized in massive data centers provided by web-scale providers. While these services will prevail for the foreseeable future, the proliferation of connected devices and digital services is driving the need for edge computing, which 
toperates outside data-center environments with distributed, as opposed to centralized, architectures.

In principle, the expansion of edge computing should benefit network operators, whose distributed network real estate is well suited to host edge computing infrastructure. However, the story is not that simple and depends on where high value edge computing functionality is ultimately deployed. 

Rather than defining a distinct edge, as the name suggests, edge computing can be implemented in many locations between data centers and end-user devices, depending on service demands, the competitive positioning of key stakeholders and the environment in which the service is being implemented. For example, autonomous vehicles and cloud RAN applications require edge solutions to be sufficiently distributed to address latency demands. This contrasts oil field implementations where edge compute is commonly campus based, somewhat autonomous and used for delay tolerant networking. 
The push for edge computing standardization

The edge compute market is still nascent, and is supported by numerous initiatives which will ultimately consolidate, and seemingly benevolent partnerships which will change as players jockey for market dominance. The telecoms industry is pursuing several edge computing standardization efforts. Notable examples include the multi-access edge computing (MEC), CUPS (Control and User Plane Separation for EPC), and CORD (Central Office Re-Architected as a Data-Center). Multi-access edge computing (MEC) and CUPS (Control and User Plane Separation for EPC) were spearheaded mobile industry. 

Although the “M” in MEC was renamed by ETSI from Mobile to Multi-Access to broaden its industry appeal, we believe MEC needs retain a mobile bias to simplify its mandate. The CORD (Central Office Re-Architected as a Data-Center) initiative aims to leverage the hundreds of thousands of telecoms central offices deployed across the globe. 

Other notable edge compute initiatives include the Open Compute and OpenFog initiatives. Open Compute provides an OpenStack feature-set extension, and is somewhat constrained by the long-term prospects for OpenStack are far from certain. OpenFog was spearheaded by Cisco with reference architectures for migrating cloud technology to the edge, with hierarchal architectures. In September 2017, MEC and OpenFog announced plans for collaboration.

Telecoms industry initiatives generally align the placement of edge compute infrastructure to the benefit of network operators. If successful, these approaches will push value back from the ecosystem peripheries towards the network. Solutions provided by web-scale providers, such as Amazon Greengrass and Microsoft Azure place proprietary edge compute functionality on enterprise managed infrastructure, which if successful will concentrate value creation at the ecosystem peripheries, akin to cloud services today. 
Knowing your strengths and weaknesses

As both telecoms and web-scale providers play to their strengths, we believe that they must also pay close attention to their weaknesses, the weaknesses of their competitors and roles of other ecosystem players. Telecoms operators are constrained by legacy operations and struggle with automation, and agile design principles that leverage contributions from open source communities. We believe that edge compute offers early mover advantages and that operators must accelerate their edge compute activities with end-users, even if pre-standard solutions are used. 

Web-scale providers are adept with agile and open source design principles but are accustomed to centralized service architectures. Short-term gains will come from their current edge compute initiatives, but only for a sub-set of services that require edge computing capabilities. 

We believe that web-scale providers will be required to evolve their edge compute initiatives, with an emphasis towards partnership strategies. In addition, both telecoms operators and web-scale providers must carefully monitor and evaluate the positioning of other ecosystem players, such as hardware platform providers like HPE and Dell and embedded system providers like ARM, Intel, Qualcomm and Western Digital. 
 There is a lot at stake as edge computing is adopted - particularly for telecoms operators and web-scale providers - and although there is uncertainty in how the market will evolve, we are certain that complacency is a bad strategy. 

Phil Marshall is Chief Research Officer of Tolaga Research




5G and data center-friendly network architectures

Matt Walker / MTN Consulting

Webscale and transmission network operators' interests are aligning as the 5G era dawns

Matt Walker / MTN Consulting

Webscale and transmission network operators' interests are aligning as the 5G era dawns

Rémy Pascal / Analysys Mason

The launch of 5G by South Korean operators serves as a first benchmark for other operators around the world