What is the State of OpenFlow at the end of 2012?
Notes for Day Three, Sunday December 17th 2012.
Today we did a lot of work with the Pronto 3290 switch.
- We used our in house tool to push flows into the Pronto until we either ran out of memory in Floodlight, or ran out of FIB space on the Pronto.
- We worked with RouteFlow via a live cd (Thanks Josh), made some modifications and constructed a standard setup for the NOX based testing.
- Constructed a script to configure the Pronto for our tests to make sure that each test is repeatable.
Some of the data that came out of our testing was the following:
The Pronto 3290, which is rated to hold up to 4k flows, held 4781, which is above the stated number. These are simple flows, source ip/port, destination ip/port.
To get this result we needed to boot the pronto into a shell instead of PicaOS. Once inside the shell we needed to configure Open vSwitch by hand, add all of the necessary ports and connect it to the controller. This is a daunting task for those who do not work with the Pronto regularly. We decided to script the setup of the Pronto and just use that each time we booted so that there were no errors. Thanks again to Josh, who created the script we modified to fit our setup.
At the end of the test we will be making all of this data public if possible (We know of no controls on the data) including a bootable CD and configurations used for the different DUTs.
We are thinking about adding Pica8’s reference setup into the test. They provide a OpenFlow 1.2 solution using RYU. You can find more information on their blog post entitled “First Integrated Open OVS Switch and OpenFlow Controller“. This would be for information only as no other vendor that we know of has a public OpenFlow 1.2 solution.