This post is a reaction to the balanced SDNCentral article by Craig Matsumoto titled “Ubuntu Has a Challenge for OpenDaylight.” In the article, Kyle MacDonald of Canonical expressed that he “likes the OpenDaylight Project, but he’s not sure he can trust it.”
I get it, and I wasn’t too far off from the same thoughts when OpenDaylight was announced last spring. A mixture of vendors that push boxes doing software-defined networking (SDN)? It was rocky at the start. ODL went through the SDN platform wars, much akin to the kernel wars of last decade.
That said, I take skepticism regarding open-source projects from vendors with a monetizable stake in the outcome with a healthy spoonful of skepticism myself. Canonical competes with Red Hat. Red Hat is invested in OpenDaylight. Got it. My message to Canonical is: Jump in and commit some code before you determine OpenDaylight is too politically charged to participate in, or bring an alternative. Competition is healthy, particularly in driving innovation. I have found OpenDaylight anything but exclusive, and there are plenty of other random community network folks I can introduce you to in the IRC channel that would say the same.
Back to ODL. We have many open-source controllers today, some from single vendors, some from research Ph.D. projects, and a couple of them are actually maintained. None of those projects is structured contractually in such a manner as ODL that expects corporations to contribute in a collective best-foot-forward by both vendors and community. That is the essence of what makes ODL unique to me and why I wholeheartedly support the project.
The Daylight Technical Steering Committee (TSC), marshaled by Dave Meyer, have managed to avoid being lured into becoming corporate obstructionist shills by their parent organizations and are laser focused on the December Hydrogen release. That is not to say some companies would like to see ODL detoured in some form or fashion, but meritocracy, community encouragement, and extreme levels of transparency create a framework of checks and balances that appears to be preserving the boundaries of corporate interests over the health of the project. This includes code committers having a path to become TSC members and even board members, regardless of ODL membership status. It feels similar to the OpenStack project technical lead (PTL) roadmap.
Code Is the Coin of the Realm
There are 13 projects being submitted for the first simultaneous release in December. Of those, three are being submitted by non-vendor contributors. While 23 percent of projects being submitted by non-vendor community sources may not sound overwhelmingly community-driven, let’s put that number in perspective. The Linux kernel, which is arguably one of the most successful open-source software projects in existence, is only around 25 percent. For a project that is only six months old, we are doing pretty well. There has yet to be a project that was denied.
Living by the credo of rough consensus and running code is tough to argue against. Your seat at the table is defined by how much quality code you commit. Peer review is highly encouraged and done respectfully, and there is a true sense of excitement at what will be in every conversation. The level of scrutiny and skepticism has actually benefitted vendors working very hard to be as open as possible. While I often compare SDN to the compute evolution, networking still has a very tight coupling of software and hardware. As a result, SDN, particularly outside of the data center, needs cooperation between vendors and community software projects.
Commoditization Is Out of the Bag
IT companies are in business to make money. Technology is consumed to save money or make money. I take vendors sniping at open-source projects with a grain of salt and understand that it’s just business, but criticism without presenting solutions, or in this case, alternatives, is uninspiring.
The differentiation of OpenDaylight comes in being under the auspices of the Linux Foundation and the immediate need for a strong open-source reference implementation of an SDN controller. Fragmentation is the only factor that can slow down SDN at this point. We are already seeing some of that. The idea of commoditizing the platform is scary for vendors that aim to lock in customer revenue. That said, it is too late; the kitten is out of the bag and will be a tiger soon enough. We have some very clear SDN architectures for the data center. What we lack are open-source implementations.
I was fortunate enough to learn from my personal hero in the networking industry, who was integral in disrupting an entire industry, that open-source software will fundamentally change networking and that we are just scratching the surface. There is far to go and plenty of time for companies to re-imprint their DNA.
When Microsoft is writing MS Office for the Android platform, the days of differentiation in the platform are long gone. My advice to those passionate about open-source and/or networking is to give OpenDaylight a spin for yourself and decide how open or closed it is. My experience has been life-changing, and I am fortunate to be enjoying it with some of the best developers in the world, along with some of the best network engineers and architects out there who are rolling up their sleeves with me to leave their mark on the industry they are so passionate about. Join us in #opendaylight on irc.freenode.net or listen to some weekly calls if you want to see what is really happening in ODL. Now get back to your IDE!
(Disclaimer: I submitted to, and am a committer on, the OVSDB project within OpenDaylight. I do not work for a vendor, nor do I get paid by an employer to work on the project. I work on it during nights and weekends. It is exciting to see I am not alone.)