WHAT
Replicate as much as possible the performance testing done in 2.8 in order to compare to 2.15

HOW

1. Run a 6 hour sustained rate test at 12 rps with a large profile
2. Run a 6 hour sustained rate test at 12 rps with a simple profile
3. Run a 1 hour peak rate test at 48 rps with a large profile
4. Run a 1 hour peak rate test at 48 rps with a simple profile

rps is users_per_second

Record parameters used in Hyperfoil. Add HTML to JIRA
We can also consider switching to Locust if reasons are strong enough

VERIFY / OBSERVE
For each test ensure that:

• Resources allocated are inline with SKU/Core/Daily API Requests recommendations and previous test runs.
• Observe total CPU and memory usage across the 3 scaling pods, backend worker, backend listener, apicast prod
• Compare totals to 2.8 totals
• Ensure that 3scale remains healthy during the tests
• No alerts firing, no pods crashing, 3scale not reporting errors (product analytics more accurate than testing tool)

References
See test results here Along with extrapolated spreadsheet with resources usage