OpenFlow News

You can now add San Diego to the list of OpenFlow tutorial sites, which includes Stanford, Brazil, and Seattle.

The latest one was held at the 8th GENI Engineering Conference, right by the beach. The weather was cool and cloudy, but the turnout was great: 58 registered and the room was filled with a mix of students, commercial researchers, and others looking to learn.  The material alternated between lectures and hands-on exercises, where attendees turned a provided NOX hub controller into a flow-accelerated learning switch.  The lecture side was given by Brandon Heller from Stanford and Srini Seetharaman from Deutsche Telekom Labs, while the hands-on side was helped by almost ten Stanford students/affiliates, ensuring that questions were answered promptly.  Slides, exercises, and more are available at the links below:

• Tutorial Slides
• Tutorial Exercises
• Longer tutorial description
• Video to come later

The next tutorial will be in Mountain View, CA as part of HOT Interconnects on August 17 – click here for details, and make sure to register by Aug 1.

OFTest is a Python based validation framework for testing OpenFlow switches.  It is intended to complement and eventually replace the current Perl “blackbox” tests which are distributed with the OpenFlow reference release.

A preliminary release of the OFTest framework, code named dragonfly-1.0, is now available.  It can be downloaded with git via the command:  git  clone git://gitosis.stanford.edu/oftest.git.  Tar files of the core and pre-built directories are available at http://www.openflow.org/oftest.  The test framework is currently targeted to 1.0 but will be updated for 1.1 as that standard evolves.

You can subscribe to the OFTest email list which provides community support for the project at https://mailman.stanford.edu/mailman/listinfo/openflow-testing.

Various documents related to OFTest are available on the OpenFlow wiki.  The read-me file is available at OFTestReadme.  A tutorial is in the works and can be seen at OFTestTutorial.  A list of currently implemented test cases and planned test coverage is at OFTestListPage.  This is an open source project and contributions in the way of test cases and framework enhancements from the OpenFlow community are strong encouraged.

Today was the demo session at the GENI Engineering Conference in San Diego and the demos included a number of OpenFlow systems.

OpenFlow at GEC8

• Integrated Control Framework Demo by a joint team of Stanford University and BBN. Using the OMNI command line tool, a researcher can reserve both PlanetLab compute nodes as well as an OpenFlow based networking substrate. The demo used the Expedient aggregate manager for OpenFlow Networks as well as the Opt-In manager. Essentially all of this demo came together over the past 4 weeks due to a heroic effort of the Stanford and BBN teams. Wiki page with more information is here.
• Expedient, a control framework with a graphical UI for OpenFlow based resources. The version demonstrated additionally can be accessed via the GENI API through a proxy.
• Aster*x, the OpenFlow based load balancer. This is the successor to the plug-n-serve system and the demo ran across a number of OpenFlow networks including Stanford, BBN, Princeton, Indiana and University of Washington.
• Transport and Aggregation. This was a combination of the aggregation demo from SIGCOMM 2009 and the optical transport integration done together with Ciena. Details here.
• WiMax. A demo from the OpenRoads team done together with two other WiMax demos at the conference.
• Clemson University showed their graphical UI for configuring the slices on their local OpenFlow deployment. The UI looked great and there are a number of similarities with the Expedient UI.

Thanks to the 20+ people involved in putting these demos together, they were a big success. A few pictures below, more in the photo gallery of the demo session.

Prof Nick McKeown recently presented a talk on Software Defined Networks (SDN) at GigaOM Network’s Structure 2010 conference. One main thesis of SDN is the separation of control logic from forwarding logic; control logic running externally to the high-speed forwarding logic makes the decisions about how and where network traffic is forwarded. This allows new services and routing algorithms to be deployed by upgrading only the control software; the forwarding elements can remain unchanged.

A video of Nick’s presentation can be seen by viewing the rest of this entry.

(more…)

The Pronto 3290 is an Ethernet switch with 48 gigabit ports and four 10-gigabit ports based on Broadcom silicon.  It supports the OpenFlow Switching 1.0 protocol with hardware acceleration for full matching support, all required actions and additional optional actions.  Pricing, likely to be under $3000 for academic institutions, and availability have just been announced to resellers.  This switch will run the Stanford Indigo software reference code which is now in beta release.  Please see the Pronto reseller page for contact information to find out details about pricing and availability.

The OpenFlow switching software release has been pushed to beta.  See the original post for more information.  Thanks to Masa, Srini and the full Stanford deployment team for integrating the development platform into the local OpenFlow network.

The first international OpenFlow tutorial was held on 5/27 in Gramado, a small mountain town in southern Brazil.  The full-day tutorial, along with a talk on Software-Defined Networking and demo session the next day, were part of SBRC 2010, Brazil’s largest conference for networking researchers (~800 were in attendance).

All events went swimmingly, with about 40 attendees.  The tutorial alternated between hands-on exercises and lecture, with topics such as OpenFlow motivation, history, interface, demos, availability, deployments, and community.  In the hands-on portion, each participant worked to learn the OpenFlow toolset and turn the provided hub controller into a flow-based switch. Fortunately, the 6499 miles and 200+ms RTT between Stanford and Gramado affected only one demo out of 6; most worked without a hitch.

The tutorial slide deck is available.

News Flash:  The Pronto 3240 and 3290 48 port gigabit switches have been announced.  These switches run the Stanford Indigo software release and are available at a very competitive price point from resellers in the Bay Area.  These are excellent candidates for those interested in doing OpenFlow research or trial deployments.

In GEC7 held in March 16-18 at Duke University, BBN/GPO and six universities (Clemson, Georgia Tech, Indiana, Rutgers, Wisconsin and University of Washington) came together and each successfully showed their current deployment status, live. All those participants have deployed an OpenFlow network in a small scale and virtualized it into multiple slices including experimental and production slices. Some universities showed interesting research experiments such as access control and mobility on the experimental slice.

It was great to see the progress in OpenFlow deployment in each campus as well as the interesting researches happening on top of it. We together will continue expanding OpenFlow deployment including Internet2 and NLR.

In a recent blog post (Will OpenFlow Software Unlock the Switch Market?), Gary Lee of Fulcrum Microsystems considers the implications of OpenFlow upon the switch market. The post notes that OpenFlow “breaks the software lock that some switch vendors have with their system customers”. Indeed, OpenFlow enables companies to develop controllers without developing their own hardware; controllers will interoperate with OpenFlow-enabled switches from any hardware vendor. We (the OpenFlow team at Stanford) expect that this will have a dramatic effect upon innovation as it enables companies to much more easily deploy new services within their networks.

We’re glad to see the increasing interest being shown in OpenFlow by industry, and we eagerly look forward to the availability of more OpenFlow-enabled hardware.