I recently watched a video of a lecture given at Stanford University by a prof from Berkeley (Go Figure). The lecture was about the definition of Software Defined Networking (SDN) and how the original team had succeeded in some areas and failed in others. The part that resonated with me was how the speaker (Scott Shenker) had dissected the networking problem.
Find the original video here: http://www.youtube.com/watch?v=WabdXYzCAOU
Every software designer that is presented with a complex problem tries to break it down into manageable chunks. I have been contemplating the application of SDN for several months and I realized that the part that would be responsible for path calculation was going to require some serious thought, algorithms and Math.
Scott Shenker arrived at the same conclusion, but he presents a clear and concise description of the problem with solid reasons for defining a new layer that he calls a network hypervisor. I don’t know that I would have reused the word “Hypervisor” but the functions that he has identified make complete sense to me.
The real question is how networking companies with respond to his analysis. The ONF has been working on an open source controller for some time, but I am not aware of anyone working on a network hypervisor. I suspect that in the time between now and the time when a network hypervisor exists, there will be a number of interim solutions.
Watch the video and let me know what you think…
Scott’s suggestions represent a significant amount of work, but that’s usually what happens when someone identifies a logical solution to an existing problem. There will no doubt be people who want to run with what we have, but it has been my experience that it is always best to bite the bullet and build solutions that last.