Since RHEL62 is no longer supported, let's look into moving the microbenchmarks over to Amazon Linux 2, to be consistent with sys-perf.

A couple of areas of known work:

1. Change performance.yml on master to take out any commands specific to our CBI machines (e.g. numactl). Change the instance type to be consistent with sys-perf tasks.
2. Once the above is successful, Run a few patch builds and see if the throughput are within 5% between runs. If not, let's sync up with build to see if we can get dedicated instances in AWS. The results for sys-perf show promise.