While the wireless industry prides itself on innovation, the telecom industry in general is not known for its speed in getting things done and adopting new technologies, but the move to SDN and NFV may be signaling an end to the days of old.
AT&T’s goal of virtualizing 75% of its network by 2020 is well-known, and Verizon is on the SDN/NFV bandwagon as well, albeit less public about its internal goals. Sprint and T-Mobile US have been less vocal about their plans for SDN/NFV, but everyone is expected to deploy SDN and NFV to some extent or another on the way to 5G.
Intel started efforts in earnest in 2009, mainly with Korean operators that were far ahead of the pack in terms accepting the idea of using server-based technology and virtualizing a lot of the applications that sat in fixed function network boxes, according to Sandra Rivera, corporate VP and general manager of the Network Platforms Group at Intel.
In 2010, China Mobile kicked off an industrywide initiative around cloud RAN, with the same idea of being able to use server-based technology and virtualization, which led to a couple of years of early feasibility work and lab trials. In the fall of 2012, ETSI launched the NFV Working Group and the first white paper was released in October 2012 describing the desire on the part of more than a dozen operators to pursue NFV as distinct from cloud/SDN. In the ensuing years, early trials and deployments occurred, with a lot of porting of applications from physical appliances into virtualized servers.
Today, almost every service provider has some form of virtualized core/Evolved Packet Core (EPC) and almost every vendor has shipped a commercial virtualization product. Commercial solutions now are moving to the edges rather than just the core.
“We really start to turn the corner at the end of 2015 and of course all of last year where we started to see the rollout of a lot of these commercial solutions in live networks across the globe,” Rivera said.
“We’ve been at this for quite some time. We see 2017 as an inflection point where you just start to see a much, much higher rate of commercial deployment of these NFV, SDN—it’s much more programmable solutions out in the market, not with just the Tier Ones but also with the Tier Twos and Tier Threes across the globe.”
All of that is foundational to 5G, with 5G acting as an accelerant for a lot of the service providers that want to move faster. If a service provider is trying to support the broad range of use cases that will be brought forward with 5G, like the ultra-low latency required for autonomous cars or remote telesurgery all the way to the low data rate required for the massive number of sensors in a smart city—all of that requires a great deal of agility.
“When you have a very broad range of usage models, you want a network infrastructure that can support it, but you don’t want to over-dimension the infrastructure for the highest bandwidth and the lowest latency types of use case because the capital expense of over-engineering for 100% peak traffic of the most stringent use case is not cost feasible,” Rivera said.
“What we’ve been doing is deconstructing a lot of that” into logical slices that are programmable, running on standard server hardware so that operators can compose in real time all the network elements they need without having to over-invest or under-invest.
Intel believes it has a unique opportunity because it offers an end-to-end solution for 5G. It has modem technology that goes into devices including smartphones, tablets, drones, cars and more, and it’s in the infrastructure, including base stations, as well as the edge and the core, where a lot of the switching, routing and authentication functions are done. It also powers the cloud and data center, so “it really is end-to-end,” she said.