For various reasons, the Arista testing started with POX and ended with RYU controllers running the switches.

The server sdn-a4, which runs the controllers, had been left in an active state with the POX controllers running in a screen session for the duration of the testing.

The first set of tests was just load testing and ran without a problem.  I was unable to saturate the switches, even when sending as many as 6 files concurrently across one switch to various destinations.

Unfortunately the screen session running POX froze up and had to be killed out.  When it was revived, it refused to allow traffic across all of the ports.  It was speculated that this had something to do with the ARP tables filling up or being depopulated.

As a result of this traffic blockage, it was decided that we would repurpose the POX code for the RYU controllers and use RYU to load multiple rule sets onto the switches for the second stage of testing.

The raw data showed that sending 6 files across one switch with 700 rules per second came close to saturating the switches (with an average of 10G/second of traffic), but did not overload the CPUs of the switches, and they handled it with only a barely perceptible slowdown.