Timeline for NFV Still Hazy
In all the hullabaloo around software-defined networking (SDN) and network functions virtualization (NFV), with new "virtualized" telco products announced almost daily, it's chastening to note that virtually nothing virtual has actually been deployed by telcos to date, and that the timeline for this much-anticipated revolution remains, for now, very much a matter of opinion. While the logic behind NFV and SDN is unassailable, and the destination unarguable, at least for leading operators, that doesn't mean they have a clear view yet of how to get from A to B, or of how long it will take.
For the time being, and probably for most of the next year, "proof of concept" will be what it's all about for the majority, and the gains from virtualization in this interim phase will largely be tactical, entailing point solutions for a few of the most easily convertible functions, such as optimization or firewall software.
Getting beyond that and applying SDN and NFV to the rest of the network -- and to the OSS that goes with it -- is a different matter altogether. At last month's big SDN & OpenFlow World Congress in Germany, for example, one of the world's leading SDN and NFV evangelists, Axel Clauberg of Deutsche Telekom AG (NYSE: DT), noted bluntly that "There is no carrier-grade SDN today." Clauberg's team is charging ahead with SDN in its TeraStream architecture, but as for the legacy DT network, "This is an architecture for the end of the decade," he said.
It's easy to see why: There are some big holes in carrier SDN that still need filling, ETSI's NFV effort remains a work in progress, and there are some major controversies still to be resolved. Should the southbound interface between control and data layers focus on OpenFlow, as the ONF hopes, or is the more hybrid environment advocated by Open Daylight and some carriers the way to go? How far can carriers use open-source approaches -- especially OpenStack -- and should they opt to use some proprietary protocols and APIs to fill in gaps, at least in the short term? What happens to the existing standards environment and its interfaces? What can be shifted onto generic COTS hardware, and what must remain the domain of specialized hardware and chipsets? None of these issues has a clear answer yet, and there are plenty of others we could add.
If there was one message from the SDN event, though, it was that the debate has moved far beyond hardware economics to a much more ambitious vision of a radically simpler network environment that will, if deployed, transform the telco industry. For DT, the aim is a network "running on autopilot" like a modern airliner. For Telefónica SA (NYSE: TEF), likewise, it's about a network that's zero-touch and fully programmable. Most of all, it's about a complete reworking of OSS.
That will be a mammoth task for the average legacy telco. It will likely require a new type of organization with a different set of skills more akin to those needed in enterprise IT. But as one operator put it, "Only a few operators see radical simplification as the aim [at present]. The rest are still watching and waiting."
In that phrase lies the main dilemma for the supply industry. It's hardly an ideal situation for strapped-for-cash vendors trying to invest and plan for the supposed revolution. Jump too early, and they risk creating a large hole in their revenue lines for the next two to three years. Jump too late, and rivals could already be embedded at leading operators before they have a product to sell.
For the big vendors, the aim must be both to provide the proofs of concept and point solutions that will dominate the first phase through the next year or two and to have a clear roadmap in place to achieve the more radical vision being debated right now in more and more big telco boardrooms.
At Heavy Reading we'll continue to track opinion through our ongoing program of online network operator surveys and interviews, and hope to be able to report back with a firmer timeline soon. Watch this space.
— Graham Finnie, Chief Analyst, Heavy Reading