In a recent interview, Google’s Head of Datacenter Fabric and Inter-DC Networking, Vijoy Pandey, urged service providers to stop waiting for SDN/NFV standards to emerge. Instead, he advised, “Just go ahead and build this stuff… start deploying a few things, and you’ll see the benefits right away.” Well, I couldn’t agree more.
The words SDN and NFV have been buzzing through service provider corridors for a few years now, but actual deployments have come all too slowly. Why? Vijoy is right in saying that many service providers are waiting for standards to be formalized. Some are still not convinced they will see enough value from it. Others see the potential, but feel daunted by the scale of the endeavor. Revolutions are scary.
Multilayer control (MLC) can turn the road to SDN into an evolutionary path where service providers can reap benefits every step of the way without diving too deep, too soon. The very first step actually happens before any SDN component is added to the network. The MLC platform can connect to legacy network equipment through its EMS or NMS and deliver SDN-type benefits, for example cross-layer network visibility and updated shared-risk reports. In the next step, SDN controllers can be added to the IP/MPLS layer only, and enable multilayer apps to operate closer to real time and perform a broader set of activities. In the final step, once the optical SDN controller is added, MLC can actively manage optical connections and add optimization and restoration capabilities well beyond the standalone value that such a controller can provide.
Other than making the road to SDN smoother and more manageable, this gradual, MLC-powered SDN deployment protects the service provider from market uncertainties. Replacing legacy gear with SDN gear only affects the interface from the gear to the controllers. Moving from controlling the network via its legacy EMS/NMS to controlling it via an SDN controller will entail some work on the MLC interface drivers (in charge of adapting vendor-specific interfaces to a common information model), but it will not impact the overall MLC platform, its apps, and northbound interfaces to higher layer systems such as the OSS or service orchestration systems. Replacing a proprietary interface to a controller with a future standard interface will have a similar localized impact. This means that the MLC platform acts as a “shock absorber” for the evolutionary process.
With this approach service providers don’t have to wait for standards to solidify. Those who are skeptical about its value will be able to see it early on, without taking risks and without installing any SDN controllers. Those who are concerned about the scope of the transition will be able to ease themselves into it one step at a time.