Big Switch Networks, one of the visionary leaders in the software-defined networking (SDN) vendor community, has stepped down from the OpenDaylight project, which aims to build an open source controller and APIs for software defined networks. Why is Big Switch leaving? That depends on how you look at it.
One way to view it is that they were squeezed out of the project by competing interests--that is, Cisco. Frankly, I'm not sure how else to view it. Big Switch joined a bit late and pushed to have Floodlight be a significant part of the controller code base.
Cisco also contributed code from its own ONE Controller product. A proposal was made to merge the two, but the OpenDaylight Project (ODL) voted to go with the Cisco controller code as a base instead of merging the Floodlight and Cisco code together. The message to Big Switch seemed to be, "Thanks but no thanks."
Now, there might well be legitimate technical reasons for this. Or procedural ones. Or practical ones. I didn't dig through all of the concerns raised that lead to a Cisco controller code base for ODL. Perhaps the ODL voting entities acted altruistically in what they believe to be the best way forward for SDN. But you have to wonder, with all the heavy-hitting big-name networking vendors in the mix, is altruism even possible?
[ For more analysis of the OpenDaylight Project, read Cisco and OpenDaylight: The SDN Application Land Grab ]
Over lunch recently, I heard ODL described by someone not all that engaged in SDN as "that Cisco-IBM project." Wow. Take that in for a moment. For an open-source project, that sort of a description is a blow. Can you blame Big Switch (which admittedly has a business of its own to build) for turning a jaundiced eye toward ODL?
Guido Appenzeller, Big Switch's CEO, highlights the company's viewpoint in his blog post on this topic.
He wrote, "Specifically, the market is clamoring for a transition toward 'bare metal switches,' or 'white box switches,' which provide customers an ability to rack-n-stack switches and centrally provision them just like they do with data center rack servers today. This market trend has recently been validated by the user community, including the recently announced Open Compute Project's networking reference design, the Open Network Foundation's ongoing OpenFlow work, as well as some other exciting projects that are forming and will be announced soon."
He also took a shot at Cisco with this line about commodity network gear: "While forward-thinking customers are committed to this trend, we question whether or not the incumbent hardware vendors leading OpenDaylight have a similar commitment."
Cisco is not at all interested in a marketplace of white-box switches because it would upend its multibillion-dollar business selling switches built on custom ASICs. Those ASCIs are one of Cisco's differentiators and a big part of its value proposition. If Cisco were to go in the white-box direction, they end up in a game of volume instead of margin, and that would hurt the bottom line. In the Cisco lingo, hardware matters. Expect to be getting that message loud and clear in the coming months.
How much of this factored into the ODL decision to go with the Cisco controller code? I'm not sure, but I guess that it was a meaningful part of what is surely a more complicated story, especially considering the number of vendors involved.
So what does this mean for SDN in the long term? A few thoughts come to mind.
First, Big Switch has a significant ecosystem built around Floodlight, and the company is banking its long-term relevance on its existing community and on continued development. In the same blog post referenced above, Guido states, "...we will focus our efforts on the Floodlight user community. Expect some exciting announcements later this year." Floodlight still matters, and it is not a foregone conclusion that ODL will eclipse Floodlight. Ah, capitalism--let the marketplace sort it out.
Second, there's room for more than one controller in the market, as evidenced by the number of open source and other controllers that already exist. ODL is not a standard, nor is ODL compliance yet "a thing." Big Switch has lots of space here to make a go of it. Conversely, ODL has lots of space to screw up, squabble internally and fail to deliver on market expectations. I doubt that's the ODL future, but time will tell.
Finally, despite Guido's assertion of market interest in white-box switches, I think that's a myopic view. The fact is that most end users have no idea what SDN can do for them, or why they might want a white-box switch.
From years of enterprise networking experience I know that many customers don't trust white-box solutions. They place their confidence in a brand that's got a toll-free support number attached. White-box switches appeal to the massive service providers of the world, but enterprise customers are also going to have their say in how SDN shakes out. Why? Enterprises as an aggregate represent more money.