Big Switch Networks, a startup with everything to gain from adoption of its software defined networking (SDN) technology, has left an SDN group dominated by traditional OEMs who had much to lose, after the group went with Cisco's tech over Big Switch's as the base of an SDN spec.
Big Switch announced on Wednesday that it was leaving the OpenDaylight SDN project because the organization's board had plumped for Cisco's network controller as the base code of an open source SDN stack, rather than a building new code repository using code from Cisco, Big Switch Networks, and other contributors.
Its opinion matters, because Big Switch was spun out from the Stanford University Lab that helped create important SDN tech like the OpenFlow protocol, has taken in funding from Intel, and has been around since 2010 – an eon, in SDN-land.
In fact, if there's any company that history can say is the "first" SDN biz, then its either Big Switch Networks or VMware-backed Nicira.
Big Switch has a big problem with what happened with the spec for OpenDaylight's SDN toolkit, because instead of assembling an SDN stack on an empty plot of land (the new repo), the community has instead decided to use a scaffold built by Cisco (its Open Network Environment controller), and then fill it in with other contributions.
This is known as the "D-E Proposal", after its advocates David Erickson of Stanford and Colin Dixon of IBM. It will use Cisco's Service Abstraction Layer (SAL) and OSGi framework, married to Big Switch Networks' net-virt contribution, the OpenDaylight foundation announced on Wednesday.
"In the end, the leadership of ODL claimed 'consensus' was reached to start the project with the Cisco controller as the base repository, despite broad community advocacy to start from a neutral repository – not an incumbent vendor’s. Thinking about this long and hard, it became clear to us that this isn’t a foundation that we can build on," Big Switch Networks chief Guido Appenzeller wrote in an acidic blog post in response to the decision.
So far, so sour grapes. But a look at the listserv of OpenDaylight members does show a vigorous debate among them, with people from Stanford, Red Hat, Intel, NEC, Ericsson, and HP all raising concerns over the decision to not create a new repo and build from there.
- "A third (new) repo to house the merged controller effort, COUPLED with a commensurate initial committer list comprised of individuals spearheading said effort makes complete sense," HP director of cloud networking and SDN Sarwar Raza, writes.
- "A clear repo and a balanced slate of committers may take us as fast as we can thru the finish line... the appearance has value too! our success and communication with new developers and potential customers is much easier when the appearance of no politics, meritocracy, multi-lateral code contribution is there," Uri Elzur of Intel says.
- "It is important to do this right than rush into something that will not be right: architecture-wise and will not have larger community buy-in. I think the idea of starting a new project with a more inclusive competent group of committers with the goal of creating a unified code base is the right approach," Guru Parulkar of premier SDN research lab ONRC says.
However, none of these objecting organizations are Platinum Members of the OpenDaylight project, and that's telling. Platinum Members must contribute $1m over two years, plus 10 fulltime engineers dedicated to producing code for their organization's SDN contributions. Moreover, Platinum Membership guarantees you a seat on the OpenDaylight board.
When OpenDaylight was launched in April 2012, its Platinum Members consisted of eight traditional IT giants – Red Hat, IBM, Cisco, Ericsson, Brocade, Juniper, Microsoft, and Citrix – and one SDN minnow, Big Switch Networks. Additional board members include NEC and Fujitsu.
"I would speculate the reasons why those other vendors [HP, Intel, and so on] chose to not be Platinum Members is because their suspicions were the same as the conclusions we came to about the parochial interests of the organization," Jason Matloff, Big Switch's marketing veep, told The Register.
Big Switch Networks is clearly the odd one out, and it's no surprise to this Vulture that the governing organization of OpenDaylight went with an SDN controller that built in an abstraction layer developed by an incumbent OEM that gives legacy companies space in which to cram proprietary tech.
"They've created an abstraction on top of the abstraction," Matloff says. "They want the abstraction layer to be high enough to support their proprietary protocols ... the technical consequences are that as a developer you do not have direct OpenFlow access."
Because of the decision to use the D-E Proposal, Big Switch Networks is forsaking its place on the board and Platinum Membership, and downgrading its involvement in OpenDaylight. It will now concentrate on working with the OpenFlow protocol.
"We've consumed a crapload of resources – and this is really why we arrived at where we arrived at," Matloff says. "The whole thing is just not viable commercially – to be honest, we wasted a lot of cycles trying to build a truly merit-based process here and shame on us for believing we could achieve it."
None of the other board members of the OpenDaylight project spoke with The Register, but we did get a tight-lipped statement from the Linux Foundation, of which OpenDaylight is a "collaborative project":
As much as Big Switch Networks would like to paint this as some sort of David vs Goliath struggle, the facts simply don't support it. It's more accurate to say this is open source vs the goals of a single, for-profit startup. In this case the developer community combined technology from multiple sources (including BSN), which the company obviously didn't like. Open source is based on compromise and working together. Sometimes strong motivations and investor goals can get in the way of that.
In the spirit of open source, we fully expect BSN to honor its commitments to this project. We also hope BSN will continue to engage in the community discussions as more eyes on code and more suggestions and alternative approaches in the long run produce the strongest code. We've said many times before that the beauty of the open source software development model is the best code always wins – we never said one company had the best, entire codebase.
With software-defined networking still in its very early stages, schemes like OpenDaylight will create blueprints for how the tech is developed in the future, and though many startups are in the space, the OpenDaylight events show that when vendors want to turn their supertankers and get involved in a tech, boy they can really move. And woe betide you if you're small enough to be rocked by the waves their changes of direction create. ®