Last week I spent some time at the “Software Defined Networking 2014” conference in London. It’s a relatively small conference I would say however given the growing interest in SDN and rapid progress of the technology it’s always good to hear alternative viewpoints and experiences. And I certainly found the previous conference here in December 2013 interesting – in particular one vendor in my view using SDN as the “hammer to crack a nut“.
Cisco wasn’t present at this conference last week, so what are others saying about SDN? Here is a quick summary of my takeaways (in some cases questions raised in my mind), which I will expand on below. And let me be controversial in my summary!
(1) Negligible discussion on live SDN deployments.
(2) NFV – at least for service providers – is potentially a quicker win than SDN
(3) SDN “Washing” is alive and well
(4) Is OpenFlow more of an academic pursuit?
(5) Open Daylight excitement
(6) Negligible Discussion on “Making It Happen”
As I say, to some my statements may be controversial – let me explain!
(1) The Status of SDN
I didn’t hear anything on live SDN deployments via this conference, either from the presentations or from the people I met over coffee and lunch. I did raise the question though, and the panel did acknowledge that the technology was maturing, they were confident “it” would happen, but may take a few years. This aligns with the recent Gartner blog discussing a survey where almost 80% of respondents in their survey were either still asking “What is SDN?” or were “Thinking about it”.
(2) Network Function Virtualization (NFV)
A big takeway for me from this conference is that there seemed to be more excitement and interest around the topic of Network Function Virtualization (NFV). One carrier at the event did say they were moving ahead aggressively with NFV. Interestingly they didn’t seem to have immediate SDN plans and observed that “OpenFlow could be quite exciting in the optical network”. This aligns with last week’s Infonetics webinar entitled “NFV: An Easier Initial Target than SDN?“, which made the point that (carriers at least) are not keen on buying a whole lot of new equipment to run OpenFlow for example, to replace the well understood networking devices and protocols they have in production today.
(3) SDN “Washing”
Yes, it was almost as if “SDN” was being downplayed during this conference I thought. One vendor talked more about “software based networking” more than what we typically define as SDN. Maybe we are (using Gartner’s technology lifecycle model, their “hype cycle“) heading over the “peak of inflated explanations” and into the “trough of disillusionment”?
(4) OpenFlow – Decreasing in Relevance?
I have to say, the lack of discussion on OpenFlow was a surprise to me – there were only passing references to OpenFlow until later in the afternoon – whereas OpenFlow was broadly discussed in the December 2013 conference here. Could it be it’s losing its relevance in the SDN debate? Certainly the lead Cisco ACI has in terms of industry thought leadership, highlighted recently in a Network World survey, seems to provide additional backing to my observation here. There was a detailed presentation around OpenFlow in optical networks, later in the day, presented by a university lecturer. There was also mention of scalability challenges with OpenFlow, and some statements along the lines of “yes we support OpenFlow but let me tell you about our real approach to SDN [which isn’t OpenFlow btw]”.
(5) Open Daylight Excitement
Open Daylight was discussed as one technology with a very promising future – and the benefits of such open source collaboration were really brought out. Good presentation on this topic I thought. One to watch!
(6) Negligible Discussion on “Making It Happen”
There is a distinct difference between “waiting for it to happen” and “making it happen”. I blogged on the latter in my first SDN blog around a year ago, and it’s still true: there still remains limited discussion on “how to make SDN happen”. At the conference, one challenging example came up with multi-layer models for SDN and NFV deployment – and analyst and vendor advice to “choose different components and vendors in each layer” e.g. choose component A from Vendor A, component B from vendor B, component C from vendor C etc. Today you probably have a small number of vendors in your network (1 or 2 say), with SDN you may end up with 4 or 5 layers and vendors. This represents a significant (and potentially costly) systems integration challenge that no-one is talking about – we’ll need a true open systems approach, with standardised interfaces between each layer, to take these costs away. Imagine 4 vendors, 4 different release trains, 4 different patch streams … and think about how you will pull this together, how you will reconcile all patch releases, all roadmaps, how you will test at scale and under load what could be a unique combination for you. Today you buy an integrated solution from one or two vendors, yet tomorrow with this multi-layer model for SDN, you may be buying a systems integration and testing challenge. Food for thought that all too many vendors and commentators are missing or intentionally ignoring. This of course is where Cisco Services can help you successfully adopt SDN.
Back then to my overall conclusion, SDN is happening, but in many cases you will have to wait for it, maybe 2 or 3 years. With solutions such as Cisco OnePK available today, with the Cisco Nexus 9000 shipping today, and with Cisco ACI showing early leadership and upcoming roll-out, can you afford to give your competitors a say 2-3 year competitive advantage by “waiting for SDN” ?