googleforgames/open-match-docs

Need guide on how to identify bottlenecks

Opened this issue · 0 comments

From several discussions in slack, identifying the source of a bottleneck when running Open Match is not trivial. We should write a guide on finding and resolving bottlenecks within the system, including any of the services not receiving enough resources:

  • Frontend
  • Backend
  • Synchronizer
  • Redis
  • Evaluator
  • MMF

Additionally, the guide should go into how to use tracing to see how long various operation are taking.

I think it may be worth building a tool which can download all of the currently searching tickets via query API, and then re-create them in another system. This would allow for testing various configurations without having to using a live production service. It could also download the tickets over time, so you can generate an artificially large number of tickets based on real data.