[SERVER-1770] Gatling perf testing with Jetty 9.4.x in Puppet Server Created: 2017/04/05 Updated: 2017/06/08 Resolved: 2017/06/08
|Fix Version/s:||SERVER 5.0.0|
|Reporter:||Jeremy Barlow||Assignee:||Adrien Thebo|
|Remaining Estimate:||Not Specified|
|Time Spent:||Not Specified|
|Original Estimate:||Not Specified|
|Epic Link:||Drop Java 7|
|Sprint:||Server 2017-05-31, Server 2017-06-14|
|Release Notes:||Not Needed|
|QA Risk Assessment:||Needs Assessment|
For this ticket, we would do some preliminary perf testing around the use of Jetty 9.4.x in Puppet Server. This would involve:
For these, we could just start with some simple 2 hour runs - simulating up to 1200 agents doing perf-control medium catalog compilations like the current oss-puppetserver-latest Jenkins job does. As we get closer to the Puppet Server 5 release, we should look into some longer-duration runs, e.g., a week long run, but I think the shorter A/B comparison run for now to get a pulse on Jetty performance is good enough.
|Comment by Adrien Thebo [ 2017/06/08 ]|
Between the perf oss-stable and oss-latest jobs there was almost no performance difference. The stable job had a mean response time of 33ms with a standard deviation of 189, while the latest job had a mean response time of 35ms with a standard deviation of 196. With respect to individual requests catalog compilation was slower on oss-latest but requests like file metadata was about equivalently fast, which makes me believe that any performance issues between stable are latest are unrelated to the jetty bump.