Performance tests for contact auto-complete

[chibenwa]

• Write a gatling scenario exercising contact search
• Run it on preprod
• Compare with OpenPaaS results.

[vttranlina]

• Run it on preprod: wait for lsc script to generate sample data

[vttranlina]

Gatling scenario linagora/james-gatling#134

[quantranhong1999]

At first, I run the default AC simulation with 5k users, 3 James pods => too much AC req/s leading timeout (>60s auto-complete).

I reduced the AC throughput by pausing it every 2 seconds.

1k users, ~300 mean AC req/s, peek 500 req/s (enabled glowroot)

Please open the following file: /home/hqtran/repo/james-gatling/target/gatling/tmailcontactautocompletesimulation-20220428053320298/index.html

1k5 users, mean 380 req/s, peek 600 req/s (enabled glowroot)

Started to have few requests timeout.

Please open the following file: /home/hqtran/repo/james-gatling/target/gatling/tmailcontactautocompletesimulation-20220428073125494/index.html

Will disable glowroot to see the impact.

[Arsnael]

Indeed... I just rechecked the simulation, no pause is too much and not really realistic either.

Wondering if shouldn't even put the pause a bit higher...

[quantranhong1999]

Indeed... I just rechecked the simulation, no pause is too much and not really realistic either.

In during section, we had been sending the auto-complete request repeatedly with no pause (up to 4 millions request in 200 seconds)

[Arsnael]

Yes we should add a pause :)

[quantranhong1999]

Yes we should add a pause :)

linagora/james-gatling#137

[quantranhong1999]

1k2 users, mean ~350 req/s, peek ~600 req/s disabled glowroot, warmed up

Please open the following file: /home/hqtran/repo/james-gatling/target/gatling/tmailcontactautocompletesimulation-20220428085742516/index.html

I think we can say our Auto-complete is at 350 request/s.

[vttranlina]

1k2 users, mean ~350 req/s, peek ~600 req/s disabled glowroot, warmed up

How many contacts?