Ericsson Research has been working to support MPLS with OpenFlow. Since OpenFlow currently lacks the features needed to describe the functionality of an MPLS switch, we proposed a set of revisions to the OpenFlow specification. We have implemented these changes in version 0.8.9r2 of the OpenFlow reference system and have made them publicly available. A version of the NetFPGA OpenFlow switch with these changes is also under development and will be available soon. More information about our project, our additions to OpenFlow, and the source code can be found on the Project Page on the OpenFlow Wiki.
Archive for August, 2009
OpenFlow for the second time in a row won the best demo award as well as the best poster award at SIGCOMM 2009 in Barcelona. The OpenFlow “Room” at SIGCOMM showcased the use of the FlowVisor to divide partition the stanford production network at Stanford into 5 virtual slices and production traffic as well as 4 separate experiments on these slices. Each experiment had its own OpenFlow controller, and the FlowVisor segregated traffic as well as forwarding decisions betweek the slices.
The individual experiments that ran in parallel were:
- OpenPipes, a system to prototype high-speed networking hardware by connecting separate components via an OpenFlow network.
- Plug-n-Serve, a controller that takes a network of switches and turns them into a distributed load balancer.
- OpenRoads, a demonstration of how multiple virtual mobility managers can coexist on a wireless network and increase transmission quality via bi-casting.
- Aggregation, a demonstration of a controller and UI that can aggregate large numbers of flows into classes of traffic and manage them efficiently.
OpenRoads was separately named the best poster of SIGCOMM 2009.
Congratulations to everyone on the team including Rob, Glen, Masayoshi, KK, Jad, Yiannis, Nikhil, Srini, Tatsuya, Guru, Guido and Nick here in Barcelona and David, David, Mikio, Julius, Adam, Brandon and Paul in the US. Also many thanks to the teams from NEC, HP and Nicira, without them this would not have been possible.
Over the next weeks we will post videos of the demos as well as the posters itself on this blog. Stay tuned!
In two weeks there will be a workshop at Stanford University on OpenFlow and how it is enabling network innovations. We will have talks on the state of OpenFlow and the NOX controller and how to create new network capabilities. You will also get to see several demonstrations of new capabilities enabled by OpenFlow in our network. Finally you will get to hear two perspective talks: Google’s perspective on OpenFlow and its implications and how OpenFlow can be a networking substrate for NSF funded nation-wide Internet research facility called GENI. We expect the workshop to be an important and fun OpenFlow event with participation from researchers from universities and industry labs, OpenFlow switch vendors and a select network providers.
The workshop will be held at Stanford on Thursday August 27th from 8:45 – 12noon in Gates B01. The workshop flier can be found here.
Update: If you are interested in attending, please do so. The only thing we ask is that you rsvp to email@example.com so we can plan accordingly. There is no fee for attending. Parking information can be found here, the closest pay parking is at the Cantor Arts Center (aka Stanford Museum).
Today, we’re announcing the alpha release of an OpenFlow interface for the Click modular router. Click is a software router that provides a rich interface for interconnecting arbitrary packet processing modules called elements. Due to years of development, Click has a large, high performance repository of elements that implement various network functionality. By adding an OpenFlow interface to Click, we hope to bridge OpenFlow’s forwarding control mechanisms with Clicks per-packet processing capabilities.
We imported the OpenFlow linux reference implementation into Click as an element – Openflow Click – combining the features of both software. This element should help researchers to build software routers that use elements from Click’s huge element library and are programmable by an Openflow controller.
This work was done by Yogesh Mundada while interning at Deutsche Telekom Inc. R&D Lab USA, with guidance from Rob Sherwood and Nick Feamster. While general thanks go to the OpenFlow community mailing list for their help, a special mention goes to Ben Pfaff and Kok-Kiong Yap for their tips and hints about understanding/using linux reference implementation of the switch and OpenFlowVMS respectively.