In an effort to solve RDS slowness IOPS was switched to PIOPS 7000, this may be much higher than we need.

https://github.com/mozilla-platform-ops/devservices-aws/pull/113/files#diff-aac6aeeaf82f5228af02042b7e7825d2R149

Plus, innodb_io_capacity, innodb_io_capacity_max also increased.

https://github.com/mozilla-platform-ops/devservices-aws/commit/69ab6c7d210e845ee735347cfdcbb42e0f86f5ba#diff-aac6aeeaf82f5228af02042b7e7825d2R68

It would be nice to know how these numbers relate; they probably do not measure the same thing.

1. Set innodb_io_capacity_max to innodb_io_capacity; reduce innodb_io_capacity to the point we see IO being constrained. As a side effect, our transactions should be getting longer
2. Adjust PIOPS downward until we see its constraining effect on IO.

This should provide us with a rough ratio, R, we can use to set PIOPS in production: Set PIOPS >= R * innodb_io_capacity_max so we have spare capacity for other processes on the machine (like the ability to login).