Two weeks ago we had a great demo of OpenFlow at the third GENI Engineering Conference. Things we demonstrated included:

• A centrally controlled OpenFlow network with OpenFlow switches deployed at Stanford, Internet2 and JGN2plus in Japan.
• Virtual machine mobility at Stanford. You can see this in detail in the SIGCOMM Demo Video.
• Flow Dragging. David Underhill created a fantastic UI that allows you to change the path packets take in the network by dragging the flow with the mouse to new routers (we have an unedited video of this part).
• Virtual machine mobility within JGN2plus and between Stanford and JGN2plus. A running virtual machine was migrated across the Pacific while hosts in Japan were communicating with it. The combination of OpenFlow and our controller allowed the virtual machine to change locations and maintain connectivity without changing IP address.

The demonstration OpenFlow network incorporated switches from (in alphabetical order) Cisco, HP, Juniper and NEC.

The slides for Nick’s talk before the demo are online here.

Thanks to Glen who was the technical lead on this demo, as well as to everyone else on the 30 person team from Stanford, HP, NEC, Internet2, Cisco and Juniper who made this a success.

More press: Cracking Open Internet Hardware on Technology Review’s web site.  We haven’t seen any switches look like the one pictured… yet.

Tim Greene from Computerworld has a very nice article about OpenFlow, vendors that have implemented it and the demo at the GENI Engineering Conference. It is also up on Networkworld.

The GENI Demo just happened a few minutes ago, and it safe to say it was a huge success. We demostrated both virtual machine mobility as well as arbitrary flow routing. More exciting updates on OpenFlow coming soon.

Update: International coverage of OpenFlow in Japanese, Portuguese, Italias, Spanish, Polish and Swedish after the jump.

Read the rest of this entry »

Last August at SIGCOMM 2008, we presented a demo of OpenFlow where both game servers and clients could move around in a production network, while maintaining their connectivity to each other.

For anyone who missed the demo, or wants to share it with others, we’ve made a self-contained four-minute video that explains OpenFlow background and the demo - with slides, pictures, and video clips of those playing the game.  Check it out above, in web-size, or see it in 720p HD!

NEC today announced their participation in the GEC3 OpenFlow demo, and their intention to support OpenFlow going forward. Press releases are online in English and Japanese. The GEC OpenFlow demo includes several NEC Switches at Stanford as well as in JGN2plus in Tokyo. NEC has been instrumental in working with Stanford to make this demo happen across continents.

The NEC IP8800 switches used in the demo are some of the best OpenFlow hardware that currently exists. They offer a good fan-out and are able to do OpenFlow at line rates. The only bad news is that what we are using are prototypes and they are not available commercially. However to quote from the release:

“NEC plans to support OpenFlow features in its future products and  to encourage universities, public testbed projects, and researchers to  experiment with innovative network ideas. “

So far we have had fairly few details about OpenFlow capable hardware on this blog, but expect this to change in the future. If you are interested in OpenFlow hardware, feel free to get in touch with us.

We are getting ready for the OpenFlow demo at the Geni Engineering Conference 3. We had a first dry run yesterday which by and large went fine. Stanford also today did a press release on OpenFlow and the GEC Demo. Full text after the jump.

Read the rest of this entry »

Yesterday OpenFlow was demonstrated at the Cisco Distinguished Engineers Meeting in Sunnyvale. We demonstrated virtual machine mobility and the graphical user interface for flow dragging. Our OpenFlow network included nodes in Internet2 and JGN2plus (Japan). The demo was very well received.

This was also the warm-up for the big demo at the GENI engineering conference next week at HP Labs. Expect to read more here in a few days.

OpenFlow version 0.8.2 has been released.  The reference implementations, tests, and tools are now available on the downloads page, and the corresponding spec is on the documents page.

Notable in this release:

• Combined package containing the Linux Reference System, NetFPGA Reference System, Reference Regression Tests, and OpenFlow Wireshark Dissector
• NetFPGA hardware accelerated forwarding (4×1Gb Ethernet OpenFlow Switch)
• Highly tested - this is the version that has been used for our 2008 SIGCOMM Demo and the upcoming GEC Demo

Enjoy!

Rob Sherwood

Srini Seetharaman

The group of people developing OpenFlow is growing and we are happy to welcome Rob Sherwood and Srinivasan Seetharaman to the team. Both hail from Deutsche Telekom Laboratories, and are working at Stanford as visiting researchers and part of the OpenFlow team.

Srini holds a Ph.D.from the Georgia Institute of Technology and brings in expertise on cross-layer interaction in overlay networks. Rob completed his Ph.D. at the University of Maryland (College Park) where he worked on network security and Internet measurement.

Expect to read more about them on this blog soon. In the mean time welcome Rob and Srini!

The OpenFlow v0.8.1 spec was released way back in May of 2008.  Since then, a number of vendors have joined the OpenFlow Consortium.  These implementations have identified a number of areas where the current OpenFlow spec is underspecified, inconsistent, or missing useful features.

Thanks to Justin, Ben and Brandon, we are nearing the release of OpenFlow 0.9, which will fix many bugs and provide a number of new features. Our hope is that OpenFlow 0.9 will contain all the features desired for the big OpenFlow 1.0 release, and only small changes will be described in 0.9.X spec releases.

Some of the most important new features include:

• Vendor extensions. This is necessary to keep 3rd party implementations of OpenFlow backwards compatible while allowing vendors to implement features they want to test.
• Better capability reporting.  Now a switch reports the features and wildcard fields it supports, for each table.
• IP subnetting support.  This enables OpenFlow switches to act as IP routers on some commodity switches.

For a more complete list of the proposed v0.9 changes, take a look at the 0.9 release notes draft on the wiki.