So, some closing thoughts on ONS. I know its a bit late, but hey, when you’re out of the office for a few days, things pile up a bit--overall, I think the ONF folks did a fine job with the event.
As I look back at ONS, I am reminded of one of my favorite IT quotes, courtesy of Bill Gates:
We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten
Long-term, I think SDN or the concepts it represents will certainly have a hand in shaping how we do networking a decade for now--how we get there and what that destination really looks like is a bit less certain.
First, I think we are early enough in the game that the technology is far from unsettled:
- Most folks are shipping 1.0 code, either literally or figuratively, and I am betting there are unseen technologies in the wings that will help shape things and I am sure folks will find interesting ways to also repurpose existing technology
- We can pretty much expect some wave of M&A to help shape the vendor and technology landscape
- As I have noted before, there is a lot of dogma about what SDN is right now that is not helpful, but I also believe it will eventually fall by the wayside
Eventually the market will sort this stuff out, and a handful of organizations are in a position to drive their own solutions, but for regular folks, I think there is enough near-term uncertainty here that it will give people pause--both in terms of customer adoption as well as ecosystem investment.
Second, the use cases are not always compelling. Most of the folks you saw at ONS this year were either service providers or acted as service providers (i.e. university IT). This is only natural, since, for these two groups, SDN is helping them do things they a) cannot currently do or b) do markedly better. On the enterprise front, I still don’t see the killer app. There are certainly some things on the horizon with potential that are related to security and on the campus edge, but, so far, I have only seen SDN as a way to do things differently, not necessarily better (at least for now). To some degree, I think this ends up being a function of the maturity of your networking operating system. For a fresh new switch vendor with limited R&D, the interesting dynamic is that the SDN movement allows you to outsource R&D back to you customers--which could become a bit of a double-edge sword for both vendor and customer. Our perspective is a bit different. While there is certainly the opportunity for our customers to customize their infrastructure, its also about making things easier to access and use. For example, one speaker referenced using OpenFlow to diagnose problems with video transmission, which I thought was a great example, but I also thought--hey, our customers can already use MediaTrace, so why re-invent the wheel?
Finally, I do not think the economic case is settled. The thing to bear in mind is that service providers operate their DCs as profit centers (its what they do), which is different from most enterprises, which operate their data centers as costs centers (support what they do). This is an important distinction, because the it guides how organizations fund and invest in their IT infrastructure--for the former you optimize you data center to maximize revenue generation capability, for the latter, you optimize to minimize expense. If you are looking to drive revenue, the only hurdle is to show you can effectively monetize the investment--give me $1.00 and I will give you a $1.50 back in a meaningful period of time. If you are looking to optimize for expenses, its a bit more challenging--while there is no inherent limit to how much revenue you can generate, this is a natural limit to how much money you can save (you cannot drive costs below 0). Often, if you have sunk cost in infrastructure that is otherwise up and running well, the best course of action is to do nothing--I find that if you are going to make an “invest-to-save” argument to a customer, you better have some solid, empirical data to back it up. The reality is that all this SDN stuff might end up being amazingly cool and useful but not really do anything meaningful for TCO. In recent history, both cloud and server virtualization were introduced with much heralded costs savings. As we have gained experience with these technologies, we have found them quite useful, but the economic angle has not always played out as expected.
All this being said, I am hugely optimistic about the future--I think there are some cool things in store--both expected and unexpected. The best advice I can offer is to stay plugged into developments, but beware the unicorn factor. Use the same critical eye you would use to evaluate any other new technology and be clear on how it can tangibly contribute to your IT exploits.