Update our load test to better match Production traffic about tokenserver

[rfk]

I'm going to focus on 997344 as the first instance of this work. @mostlygeek what's the specific concern here - is it just the relatively higher db write-load in stage, or something more? IIRC our loadtests have pretty much always been CPU-limited just like prod appears to be.

[jbonacci]

Whatever change we make to the load test , let's make sure it remains configurable.
That is, we may want to change it to better match what is currently seen in production.
But, we should be able to do the following:

1. Return to the previous load test configuration
2. Be able to add future configurations that might more closely match production as we go from quarter to quarter (especially during 1.5 migration)

[mostlygeek]

IIRC our loadtests have pretty much always been CPU-limited just like prod appears to be.

I think it's the opposite. When we were testing before we had 3x m3.medium and the limitation was the write IO on the DB server (500 IOPS). In prod we have to run 6x m3.medium and db write IOPS is between usually between 10 and 20 IOPS.

[rfk]

Closing this in favour of https://bugzilla.mozilla.org/show_bug.cgi?id=1022721