2017 Predictions: SDN in 2017 – domain-specific solutions powered by specialized processors, zero latency and open source
Source: RCRWireless News
Telecom operators are expected to continue investing and deploying in SDN platforms, though advances need to be made in order to reach their full potential.
Editor’s Note: With 2017 virtually upon us, RCR Wireless News has gathered predictions from across the mobile telecommunications space on what they expect to see in the new year.
Telecommunication operators face competitive pressures everywhere. To stay in the game, they are deploying advanced networking technologies like software-defined networking that make networks smarter and more agile. At the core of SDNs are controllers that manage flow control to enable intelligent networking – an improvement that should should help evolve operators from traditional transport to more agile cloud-like service provider. And in 2017, I expect we’ll start to see domain-specific SDNs that use specialized processors capable of delivering industrialized network virtualization with near zero latency, even for web-scale, highly distributed carrier-grade services.
Sizing the SDN market is tricky and estimates vary, but it’s already in the billions. Most market researchers expect strong growth – 50% or even more year over year – between now and 2020. Research firm IDC estimates that by 2020, the SDN market will be worth nearly $12.5 billion, and says much of that growth will come from software, specifically the virtualization/control layer and SDN applications.
I’ll add some caveats, however. First, that growth will slow if SDN runs on commodity x86-based routers and switches. Commodity network devices have reached their upper limitations in terms of performance and agility and they simply won’t scale at the rate needed for future demands. Growth will also slow if SDN doesn’t include automated orchestration, isn’t intelligent or can’t perform near zero-latency transactions. Operators also can’t simply upgrade existing networks, add more bandwidth or layer SDN on top of legacy infrastructures. It really requires SDN architectures built from the ground up on domain-specific processors in order to achieve zero-latency computing and automated services orchestration. Then, operators really will be able to move application resources – whether network, compute or storage – to wherever they are needed, at any scale.
By the way, the vision of zero-latency computing would ensure that virtually almost no time is lost during the exchange of information between one interface and another. That means a system can respond instantly to an input of information, and this is particularly important for applications such as medical robots as well as many “internet of things,” industrial virtual reality and other next-generation applications. Automated services orchestration is the automated arrangement and management of computer resources, and typically involves aligning business objectives and requests with the applications, data and infrastructure through workflows, provisioning and change management.
But let’s get back to my SDN prediction for 2017. It’s been nearly 30 years since we first saw graphics processing units – specialized electronic circuits to improve the creation of images for output to a display. GPUs are much more advanced, with highly parallel structures that make them more efficient than general purpose CPUs for algorithms. These evolved to GPU accelerators, first developed in 2007 by Nvidia, that use a GPU with a CPU to accelerate the processing of analytics, engineering and other intensive applications. GPUs are now used in data centers around the world and are also increasingly being used for applications in everything from virtual reality, to artificial intelligence, to self-driving cars and robots. In the coming year, I expect we’ll start to see domain-specific SDN applications that can leverage specialized processors similar to how GPU accelerators impacted computing.
The smartest and strongest companies will begin to move off of x86-based platforms for routing and begin investing in more advanced platforms. GPU accelerators will be able to accelerate the SDN controllers as well as accelerate the automated orchestration of network resources. The result? An intelligent, zero-latency network service that can scale to the millions.
Consider a global manufacturer with facilities spread around the world. With a virtualized SDN, the network supporting the applications is more dynamic, flexible and centrally controlled. But will it support a cloud-based model that centralizes all the intelligence needed to manipulate robots, conveyors, assembly lines and other processes at numerous locations, in the manufacturer’s corporate data center? Will it deliver that intelligence without any latency? Will it orchestrate network resources automatically and intelligently? In order to achieve this, the manufacturer needs industrial virtualization, a domain-specific SDN capable of supporting advanced applications like virtual reality and IoT among millions of possible traffic flows, and delivering it all with zero latency between the data center and all the manufacturing facilities.
In addition to my prediction that 2017 will be the year domain-specific SDN applications take shape, we’ll also see accelerated telco participation in and adoption of open source. There are a number of benefits to open source, such as the upstream communities of innovators that operators can directly participate in to help lead development of and constantly improve SDN technologies and solutions. Telcos and their partners are joining open source initiatives, including the Linux Foundation’s OpenDaylight Project, which is working on a common SDN platform and ecosystem based on open systems and open source, and the Open Platform for NFV project, as well as open source software platforms like OpenStack.
In the coming year, everyone from enterprises to service providers and carriers will set their sights on SDNs, especially open source SDNs. The SDN market is expected to be worth tens of billions of dollars in just a few short years and the technology promises more efficient, flexible, intelligent networking. But achieving large web-scale SDN will require more than virtualized networking built on top of legacy architectures and commodity x86. It will require a fundamentally new type of hardware platform, one that’s built for domain-specific solutions using highly specialized, powerful processors. We’re at the chasm, and I expect us to cross over soon.