This post is also available in: Japanese
Rarely a week goes by when we don’t speak with a senior executive or technologist at a service provider or enterprise asking us to validate or invalidate some claim made by a networking, virtualization, software-defined networking, OpenFlow, or SDN company. To save everyone some time — we decide to produce ‘SDN Myth Busters’ and share our findings with the community:
If I had a nickel for every time I’ve heard this statement — I’d be retired and making fine Bordeaux blends as a full time profession. Most network hardware vendors claim to have some level of partnership with one or more controller vendors; yet when we speak with customers — we’ve heard that just about every networking company has told them they are building their own SDN controller. This should be no surprise – what vendor is going to give up the ‘brain’ that can control their physical or virtual data plane technologies to another company? None.
Conclusion: Controller is strategic and is a strategic asset for any network or virtualization vendor which is validated by the action of partnering while building their own controller software and IP. So are Northbound APIs, more on that in a later post.
Myth #2: Open-source will save you from vendor lock in…
There’s a long history where successful open-source projects have been controlled by one vendor resulting in vendor lock-in (MYSQL and OVS to name two). I define vendor lock-in as a situation where for one reason or another, the expense to switch from one vendor to another is so great that a customer is tied to that vendor’s technology. As we look at SDN technologies, this is true for Floodlight and Indigo which are controlled by Big Switch (as is Open vSwitch from Nicira / VMware). Juniper is also pitching an open source SDN alternative. Customers should realize when they make a purchase decision that includes open source technologies which lack a broad community with a large number of independent developers — you are in the same vendor lock-in position as you are with non-open source software. (Editors Note 12/2/12: See our post on OpenSource: The Biggest Risk to SDN)
Conclusion: In networking, most cases open source locks you in similarly to non-open source software. We have a suggestion as to how to build a broad developer community for OpenFlow — though so far we’ve found no takers (even at ONF and On.Lab).
We hear this from vendors (both large and small) all the time. After speaking with end customers we’ve come to the conclusion that each vendor: a) has a handful of installations (most in proof-of-concept stage); b) each installation is for a slightly different use case; and c) most customers are not yet reference-able.
Conclusion: There are limited reference deployments today and the ones that are, are not broadly repeatable.
Myth #4: We have revenue (…followed by “we are cash flow positive”)
In addition to customers, we also speak with lots of venture capitalists — which means we hear lots of rumors, receive lots of questions, and perform due diligence on a number of SDN start-ups. Based on what we see today — most SDN start-ups have limited revenue at best — which means we have a hard time believing that any are cash flow positive today. We expect that some may have received non-recoverable engineering (NRE) income from some large networking vendors which maybe claimed as revenue (though this isn’t really true customer revenue nor a measure of traction).
Conclusion: We don’t doubt that some SDN start ups companies have been paid for services rendered, though doubt that much is from end customers. Customers, as you make supplier decisions, validate this claim by asking the start-up for the sources of this revenue to determine if it’s sustainable and if it represents real customer traction or just a partner paying for some custom engineering. We also have thoughts on the invest-ability in SDN — Nicira notwithstanding — that we believe still generally holds true today.
We hear this from everyone (new and old) and in their next breath they add ‘we’ve stolen XYZ from {fill in blank}’ where blank equals one or more of the following: Arista, Nicira, Big Switch, Cisco, Juniper, VMware, etc. While it’s impossible for everyone to have the ‘best’; there is a ton of networking and SDN talent in the market and its clear that for the first time in years there’s lots of people ‘on the move’
Conclusion: Given the maturity of networking there’s lots of great engineering talent available — meaning engineering wise, it’s true that everyone has a great engineering team. We have a few thoughts on the price to join an SDN start-up.
The best part about these myths is that they are part of what happens when a technology is being commercialized — and most of them are harmless things vendors tell early customers to get them comfortable with a purchase decision that no-one else has ever made — so our advice — is take each claim with a grain of salt and validate the claims with 3rd parties to separate the white-lies from mis-information, and repeatedly ask questions for detailed SDN customer success stories before you make the final vendor selection and write that check.
Do you agree or disagree? Comment below.
NFV and SDN: What's the Difference?
Top 5 Highlights from Open Networking Summit – ONS 2013 Tutorial Day
SDN And The Forgotten Data Plane – Is My Flow Equal To Your Flow? 
Exclusive: Shining the Spotlight on OpenDaylight-What you MUST know about the...
Featured Post: OpenStack Networking--Operational Simplicity With Cisco Nexus...