assistance with interpreting autopilot results

[arizona-1995]

hi,
I ran a workload test in hammerDB against my Oracle 19c database using the following specifics and I'm looking for assistance with interpreting the results of my autopilot test. We're migrating our database onto new servers with an upgraded Linux operating system and I'm looking to compare old with new. Here are the test attributes and results. Could you help me understand what conclusion to make based on these results?

number of warehouses: 300
ramp time: 2 mins
test time: 5 mins
autopilot sequence: 1 5 10 15 20 25 30 35 40 45 50 55 60 65 70 (each test was 10 mins)

Client OS: Windows 11
Cores: 16
Memory: 32 GB

Database OS: Linux
Oracle version: Oracle 19c
DB server cores: 4

[sm-shaw]

Note that you can click on the legend below to show both NOPM and TPM individually. What you have generated is a performance profile and are showing the performance results for one system with a peak result of 60K NOPM on your 4 core database server.

Now also do further tests at your peak result (25 Vusers) and enable metrics and/or time profiling and the transaction counter. You should see a level transaction count and stable CPU utilisation. (In the GUI start metrics and transaction count). Also view the AWR report that is generated for this run and look at the Oracle metrics as well. You can also do the same tests with a script in the CLI.

Once you have this data do the same for your new system, generate a performance profile, drill down into collecting more data at peak performance and then compare the two. You will then have a comparison for your new upgraded Linux OS and be able to determine if everything is as you expect.

[sm-shaw]

I also should have added that HammerDB has a very nice active session history viewer for you to analyse the statistics for your Oracle performance tests. In the GUI you can drag the box over a particular time period and see the statistics for that time. So as below we can see that log file sync is the main wait event here: