1 Comment

Summary:

The idea of software defined networking enabled by the open-source OpenFlow protocol is under threat from corporations intent on using the OpenFlow name and the promise of software defined networking to lock buyers into their gear, according to a Big Switch Networks executive.

Guido Appenzeller (left) and Kyle Forster of Big Switch

Guido Appenzeller (left) and Kyle Forster of Big Switch

The idea of software defined networking enabled by the open-source OpenFlow protocol is under threat from corporations intent on using the OpenFlow name and the promise of software defined networking to lock buyers into their gear, according to a Big Switch Networks executive. Kyle Forster, a co-founder of Big Switch says the company has encountered buyers of its technology that are concerned and confused about what is so open about OpenFlow and SDNs in general.

So is OpenFlow, the open source protocol that underlies software defined networking, getting subsumed by corporate interest and a marketing white wash? Or is Big Switch Networks, which is pushing an open source version of a new networking technology, promoting this idea as a means to differentiate itself and discredit the competition?

The technology we’re talking about here is so early, that products using OpenFlow or pitching software defined networks were just announced last year. And so far, few companies have put any of these products or ideas into practice in their production networks. But if you want to control a developing technology, it’s good to get in there early.

A big opportunity in networking awaits.

Big Switch is one of many startups attempting to build technologies that will virtualize the networking layer inside the data center, and it hopes to do this with an open source-based controller that can run on commodity hardware. So far it hasn’t announced clients, but Forster says folks are out there using its software. The benefits of an open software defined network setup are twofold: owners of virtualized servers can manage their networks using code instead of messing with the physical infrastructure, and they can implement this without paying big bucks for specialized gear that requires specialized code to control it.

The creation of OpenFlow, an open source protocol that allows a commodity server to tell a switch what to do, has enabled companies like Big Switch to build out controllers to offer software defined networks. However, other companies are also offering software defined networks– those that are programmable and virtualize the physical infrastructure. But those aren’t necessarily built on Open Flow.

Nicira, for example offers a way to build scale out virtualized networks and uses OpenFlow, but only as a small aspect of its controller product. Other vendors such as IBM, HP, NEC, Cisco and Juniper also are offering software defined networks that may incorporate OpenFlow, but also have other elements that are specific to that vendor and their gear. IBM and Juniper are both talking a lot about services and systems that are tied to their boxes.

Big Switch is taking a different tack, attempting to not only rely on the OpenFlow protocol, but also open sourcing its Floodlight controller, that actually controls the virtualized network. Big Switch hopes to make its money by offering consulting and services for customers, such as creating easy to build stacks of infrastructure using OpenStack and its Floodlight controller for example. Floodlight has been downloaded more than 1,000 times saus Big Switch.

Get ready for FUD, claims of closed ecosystems and OpenFlow-washing galore.

The battle here is fairly significant. Networking has to change to meet the needs of virtualized and webscale infrastructure. In many cases where compute is elastic, it’s the network holding systems administrators back — the elasticity stops at the network. But networking is also a big business for the companies mentioned above, and one they’d rather not see become commodity hardware and open source software. That could crush their margins.

Adding to the stakes is the general lack of know how about networking, which has been shrouded in speciality programming, given credit and blame for security and compliance issues, and is overall pretty complicated stuff. So the folks trying to keep their SDN-enabling products less interoperable have a good reason to do so. Making SDNs easy to deploy takes a lot of know how. But on the other side, many large-scale data centers aren’t going to settle for expensive, proprietary systems that they can’t easily adapt to their needs.

Big Switch agrees. In a release this morning it offers:

Successes such as Hadoop, MySQL and Linux demonstrate the importance of open source in every major software revolution that has taken place in the past decades. As networking is becoming more software-oriented, open source provides complete transparency on the quality of its code while enabling customers to benefit from contributions made by the active open source SDN community and more importantly prevent vendor lock-in in the new network landscape.

That touches a bit on the issue of continuing innovation, which can be hard in a closed ecosystem. Nick Lippis, industry analyst and publisher of the Lippis Report, says that Big Switch may be using the fears about OpenFlow getting closed off as marketing, but the issue is an important one if the industry wants to continue innovating. Lippis believes it will be crucial for switches to be able to talk to a variety of controllers as well as open APIs that allow programmers to talk to the overall network.

That’s the future, but already he sees worrisome tactics already at play by networking vendors. “We’re seeing more [networking] companies taking the term SDN and repackaging their products in an SDN context, Lippis said. “And the open issue will be important as we move further into product adoption.”

So the question appears to be if OpenFlow and software defined networks will be closed off before they ever really get a chance to bloom.

  1. Stacy,

    A provocative article. My reaction is too long to fit as a comment, but I posted a blog entry at cplane.net (FKA layerzngn.com) with a link to your article. What strikes me is growing, not lessening, confusion over what SDN is, whether OpenFlow is tied to it and whether OpenFlow is closed or open. Another topic for discussion is whether open source is or is not a core attribute. ONF’s no doubt well-intended efforts may be unwittingly contributing to the confusion and uncertainty.

    Harry Quackenboss (quack@cplane.net)

    Share

Comments have been disabled for this post