Investigate performance issues

[ljacqu]

from @sgdc3

https://timings.spigotmc.org/?url=jeracesopa We are still having performance issues... (timings by @hsombini)

This issue is for finding problems (improvements) regarding our performance. Once we have some figures we can create follow-up stories with concrete improvements.

1. Are we creating too many async tasks?

• Comparison async vs. non-async. Probably with @games647's LambdaAttack we can test with the same conditions.
• Comparison with 3.4 --> measurements but also code comparison
• Is it in particular sections of the code that we create too many tasks?

2. Where do we lose the most time?

• Log times to an additional file. Perhaps with https://www.infoq.com/articles/perf4j?
• If configurable, could be made available to some kind souls that would provide us with data... :)
• Analysis of results

3. Check database interactions

• Check for improvements regarding DB interaction—if we read or update data we should only do it with one method.
• Method calls to the DataSource interface can be logged with a proxy: https://github.com/ljacqu/AuthMeReloaded/commit/1599fba0d94b61d006df9c015cc0fba3e15ff044

[sgdc3]

https://timings.spigotmc.org/?url=jeracesopa We are still having performance issues... (timings by @hsombini)

That was ASYNC

This is SYNC in a similar condition: https://timings.spigotmc.org/?url=ponopubufu

[sgdc3]

Log times to an additional file. Perhaps with https://www.infoq.com/articles/perf4j?

Perf4J looks very outdated, what about that? http://metrics.dropwizard.io/3.2.0/index.html

[ljacqu]

@sgdc3 Looks nice! I'm all for it.

[DNx5]

It looks normal to me, isAuthAvailable is called when players join. we already cache the value to save performance

[sgdc3]

People are refusing to update from 3.4 to 5.2 due to performance issues, we can't ignore the problem.

[sgdc3]

I'm now personally using 5.4 on my network, never experienced lag issues caused by authme, my timings looks fine. I'll move this to 5.5