The hype about Software Defined Networking is running high, with the Open Networking Summit seeing upwards of a 1000 attendees and a large number of vendors, incumbents of the networking world and small startups alike clamoring to show off their gear. But, sitting through the sessions and walking through the vendor demos, it told me one thing – if we thought SDNs are going to bring administrative, topological or operational efficiency to networking, that is probably not going to happen.
Why am I so pessimistic about this, when the world is more or less giddy with the endless possibilities that SDNs can bring in these directions? My pessimism has nothing to do with the fundamental capabilities of the technology, but has everything to do with the fact that the space is filled with carriers (and enterprises) who want to cut costs, incumbent vendors who want to stay in business and save their falling profit margins, new vendors who want to make it big and displace the incumbents and a research community that is largely just giddy about coming out of decades of boredom that was infused by working on improvements to BGP, TCP and QoS that nobody in the real world cared about. That’s correct, nobody is really thinking about the users or about designing it right.
So, anyone that is looking to reap the benefits of the technology in any reasonable amount of time better have pockets deep enough to build it end-to-end themselves. I will refrain from making references to my employer’s SDN adoption itself. There is enough public domain information on it and for the technology enthusiast, Amin Vahdat’s talk at the ONS should provide a lot of juicy details.
Sadly, yet again, here is a technology with a lot of potential to make us rethink the way we deploy and use our networks without an ecosystem with motivations that would support that. The innumerable marketing pitches made at the ONS are a testimony to this impending future. We have seen this in the past with the cellular world. The end result of such an environment is often numerous specifications drafted with many compromises to accommodate various favorites, with very little true interoperability. Although the IETF has had better success with interoperability, it has had other issues, notably, incredibly long times to reach consensus and have a spec published. Also, there is the mess of producing large numbers of incredibly complex specs. SIP, IPv6, anyone?
Even though the SDN space is showing signs of becoming another giant mess of specs and gear, there is always the hope that people will take Nick McKeown’s talk seriously and start thinking about the core strengths of the technology. The hope is small, but all is not lost yet.
But, more importantly, I believe (and hope) that one positive thing that will come out of this SDN resolution is a massive rethinking of networking APIs and just the way the networking world approaches software. So far, it has taken people with extensive knowledge and experience in the intimate details of the vendors’ gears to be able to operate networks. That has the potential to be disrupted with the SDN wave. This won’t happen from the incumbents, but hopefully the S in SDN has attracted enough software talent to this field to cause this. Although the Internet has largely been running on software thus far, there is some feeling that SDN is bringing software to networking. Shshsh! Let’s keep it that way, in the hopes of seeing better defined programmable interfaces!