FAUCET (https://github.com/faucetsdn/faucet), originally by REANNZ, is an open source SDN/OpenFlow controller for experimental and enterprise networks. FAUCET runs in production at multiple sites and supports multiple hardware vendors. This blog describes FAUCET itself and FAUCET use cases. The blog author works at Google and has contributed to FAUCET. Any opinions stated here are my own, not those of Google.
The following shows an Agilio CX 2x10GbE card controlled by FAUCET. As you can see, because the interface to the card is just OVS, FAUCET works exactly the same way as if it were controlling an OVS software-only switch.
The 2008 publication of "OpenFlow: Enabling Innovation in Campus Networks" introduced the idea that networks (originally campus and enterprise networks) can be treated more like flexible software rather than inflexible infrastructure, allowing new network services and bug fixes to be rapidly and safely deployed.7
Since then many have shared their experiences using SDN (software-defined networking) and OpenFlow in wide area and data center networks, including at Google.10 This article returns to enterprise and campus networks, presenting an open-source SDN controller for such networks: Faucet. The Faucet controller provides a "drop-in" replacement for one of the most basic network elements—a switch—and was created to easily bring the benefits of SDN to today's typical enterprise network.5
I can’t tell you exactly where Guru and I were this week, but we were visiting one of our hyperscale member companies, talking about open source software. It was really exciting to see that some of the work we have been doing is actually being leveraged by “the big guys.” While we were talking about plugfests and programs and operations and development plans and communities and all the tools and structures to support dynamic open source projects, our host leaned over and looked at Guru and me as the conversation paused. Guru asked, “What can ONF do for your organization?”, and there was a thoughtful pause.
“Tell everyone . . . OpenFlow is not dead,” were his words, carefully delivered, as if he had been rehearsing them for some time. “Take due care of OpenFlow”, he seemed to say, as if there was some risk that in the hurtling rush to “The Next Big Thing”, we might accidentally divert our attention from some of the core plumbing in the SDN ecosystem.