Are Carriers Taking a Hand at the SDN/Cloud Boundary?

It seems the SDN scene just never goes to sleep.  Today at the SDN and OpenFlow World Congress a group of Tier One operators announced they were launching an ambitious initiative called Network Functions Virtualization (NFV) to standardize network infrastructure and virtualize network functions/services.  As I’m hearing, this activity goes beyond OpenFlow and SDN to address standardized cloud infrastructures and even a possible PaaS-like structure on which to build services.  However, it’s not clear just how far it will go in this ambitious direction, or how fast.

What seems to be the key point of NVF is the abstraction of network infrastructure functionality, something that could in theory reduce it to a model whose component elements had specific interfaces and functionality.  These elements could then be mapped to equipment with the assurance that anything that represented itself as “Element X” would perform like any other option, and so components of infrastructure could be plug-and-play.  This is clearly something vendors would hate, and just as clearly something operators would love.  I love it too because I’m a big fan of the abstraction/decomposition model in general.

The challenge here is to get the abstraction model going.  “Analysis” in logic is the process of breaking something down to constituent elements; “Synthesis” is the process of creating something by assembly.  We seem to get stuck in synthesis in networking, starting at the bottom and building up to the top.  In community activities, as this one already is, the risk is that you assemble legs and trunks and tails and never quite come up with the elephant.  If you proceed on an analytic basis, you start with the high-level goal, which makes it harder to get off-track.  Vendors typically create the synthesis pressure because they want every story to be a box story.

The thing is, there’s a lot of open real estate at the top right now and a lot of disorder at the bottom.  A vendor who can simply rise above the din and get to the value proposition has the opportunity to create all the important elements of NFV before anyone get there.  You can see a bit of this in how SDN principles fit into transport networks, a topic that was the subject of a Hangout on Google+ with Ping Pan (IETF and OpenFlow guru) and I (http://www.youtube.com/watch?v=5kSnxsYyjiY).  The same video demonstrates the “synthesis/analysis” dichotomy.  We can start talking about how to fit OpenFlow to a transport model, but we can also start by talking about what’s different in transport SDN than in “packet SDN”.  You decide which you like!

Cisco, I think, has aspirations at the higher level of SDN.  In any realistic SDN model, SDN has to start with the cloud.  That means you have to map the virtualization that SDN uses to represent the network’s services to the virtualization that the cloud recognizes.  Right now, the latter is represented in OpenStack’s Quantum interface and in a number of DevOps projects that are dealing with cloud deployment and (of necessity) how network connectivity figures into that.  If you start there, you can visualize a cloud network as a virtual structure with abstract properties, and you can visualize provisioning as an orchestrated process that assigns resources to meet those abstract goals.  NVF seems aimed at generating just this sort of thing, but is it going to do it by starting with the cloud or starting with the network?  If the latter, then Cisco may own all the useful stuff before anyone even realizes it’s needed.  Which, of course, is likely what Cisco wants.

NVF is the product of some of the largest Tier 1 providers in the world, and if they really get focused on the goal line they’ll make a major difference.  Perhaps the most significant difference is that they’re going to level the playing field in terms of the mapping of functionality to equipment, insuring that vendors are tested on common capabilities.  That shifts the notion of differentiation from box-to-box competition (which is possible only on price in an NVF world) to strategic positioning of your products to meet the functional goals NVF provides.  In other words, it’s time to get out of the bit weeds and think about clouds, network vendors!  That’s where the definitions that set your future will come from.

Leave a Reply