Some network vendors may still have dreams of keeping carrier virtualization within their proprietary environments, but one of the key reasons why operators are interested is the chance to break away from technology lock-ins and implement truly interoperable, mix-and-match networks at last. This goal was clearly seen in the new mission statement issued by the operator-driven ETSI group which has been driving the specifications for NFV (network functions virtualization), the key standard for bringing this technology to carrier environments.
The impact and success of NFV has been unexpectedly powerful, indicating that there is a pent-up demand for the benefits it promises – lower cost of ownership; an end to lock-ins; greater flexibility to allocate network resource where required and to improve quality of experience. NFV achieves this by running key functions in software, as virtual machines on off-the-shelf hardware, and in many cases may be a precursor for full SDN (software defined networking).
The intense interest in NFV has led to a broadening of the remit of ETSI’s NFV Industry Specifications Group (ISG), whose work was originally scheduled to be finished at the end of this year. However, at the seventh meeting of the ISG, it was agreed to endorse a second phase, which will look to extend the ecosystem by forging ties with open source organizations and other industries, to ensure interoperability and open standards remain at the heart of the technology.
The ISG has a new chair and vice-chair – Steven Wright, lead member of technical staff at AT&T, and Tetsuya Nakamura, senior research engineer at NTT Docomo. Both these operators, of course, tend to be heavily involved in setting standards and pushing next generation network technologies, and both have extensive trials of various aspects of virtualization and SDN.
At the meeting in Silicon Valley last month, attended by representatives from 37 operators, NFV Phase 2 was described as the implementation stage, and a new mission statement emphasized the goal of an “open, interoperable ecosystem, through specification, implementation and deployment experience”.
Don Clarke of CableLabs, chair of the NFV Network Operators Council, told the meeting that interoperability was the primary issue for service providers. He said: “In the next phase we intend to intensify our efforts towards achieving interoperability for NFV, which includes normative work where appropriate, formalized dialogue with other industry and standards organizations including open source communities, and a continued emphasis on encouraging open NFV implementations and proof of concepts.”
This reflects the broader push not just to standards but to open source building blocks to create a fully open platform to underpin the software networks of the future. This is why it is important for the ETSI NFV group to emphasize openness and remain carrier-driven, to avoid accusations that it will be overtaken by vendors wanting to put off the evil hour of full SDN. The biggest supplier proponents of NFV have often been those with the biggest hardware revenues to protect.
Though virtualization and SDN can be implemented separately, many operators see NFV as a first step towards a more completely software defined environment – and some providers clearly hope it may slow the progress to full SDN by delivering some of the benefits, such as reduced cost of ownership, without abandoning specialized boxes altogether.
Indeed Dan Pitt, executive director of the Open Networking Foundation (ONF), which supports SDN for carriers based on the OpenFlow spec, said at an event in June that NFV would become “kind of irrelevant” in the medium term, because it is too hardware focused. Though “”critically important” in the near future, NFV will be “of diminishing long term benefit, because really what it does is virtualize hardware defined networking,” Pitt said. “It will become kind of irrelevant, because it will all just become programming modules.”
Virtualization moves some or all of the functions of the network from dedicated hardware appliances to virtual machines running on generic hardware, and creates tunnels between those VMs so they can be reconfigured on the fly without affecting the whole system. SDN separates the control and data planes, centralizing all the control of the huge number of VMs machines on standard servers and allowing the behavior of the network to be freely programmable, using standard interfaces and common developer tools.
In March this year, the ONF signed a memorandum of understanding with the ETSI NFV group to explore how SDN could “enable forwarding plane support for some of the most important NFV use cases”, bringing the two separate but related approaches closer together. Pitt said at the time: “NFV does not go very far without SDN as a supporting framework. NFV does a great job of converting boxes to commodity hardware or even virtual machines, but if you’re doing forwarding you need SDN too. Without SDN many functions can’t be virtualized. You need the forwarding plane to do anything.”
While some suppliers may be branded cynical for insinuating proprietary, if programmable, hardware into the NFV and SDN platforms, in reality, most carriers will take a step-by-step approach to the migration, spread over a decade, and most will not want to leap too rapidly into a ‘pure’ approach based entirely on commodity servers and web services. For one thing, they are not yet convinced that these open platforms are yet able to handle the intensive and highly specialized activities of their networks to the level of performance and reliability afforded by dedicated platforms.
So specialized, optimized hardware will not disappear from carrier networks for many years, but increasingly it will have to be low cost, low power, simple and programmable to find its way even into conventional deployments, as the web giants and the SDN juggernaut reset expectations even among operators which have no plans to go software-only for a decade or more.