Comment 7 for bug 2024019

I did 3 runs pre/post patch and recorded the data in the attached file.

Post-patch means using my PPA from [0].

I had to set up these systems in a different AWS region. I chose eu-west-1 simply because it supports the AWS a1.metal instance type which is the first graviton instance. I set these systems up as described in[1], otherwise the test performed was identical to[2].

Overall I think the degradation in performance is nominal in this setup. Notes I took on the results:

* I saw a 0.72% increase in CPU usage
* The strex events seem relatively the same, the first run post-patch seemed a little high, but that could just be due to inconsistencies due to an e2e test like this, the 2 runs after seem to equalize.

To summarize the results in the txt file:

request/s
pre-patch: (160287+152196+154558)/3 = 155680
post-patch: (154558+162545+161283)/3 = 159462

strex events:
pre-patch: (894+867+883)/3 = 881
post-patch: (913+879+872)/3 = 888

CPU usage (only did one run, explained how this is gathered in txt document)
pre-patch: 98.77%
post-patch: 99.49%

[0] - https://launchpad.net/~mitchdz/+archive/ubuntu/lp2024019-moutline-atomic
[1] - https://github.com/mitchdz/aws-nginx-testbed/blob/main/ARMv8.0-test.bash
[2] - https://github.com/mitchdz/aws-nginx-testbed