Open theotherphil opened 1 year ago
Well, I expect that the bottleneck is clear, it's the database :) So I guess that the way to make it faster would be to use the classic tricks: adding indices, caching etc. The DB has been acting a bit strangely recently. It contains relatively large amounts of data, and there may be dragons.
We don't have any telemetry nor profiling setup at the moment, so that would also be useful to confirm my "hypothesis" :)
Can you say more about the specific request you were making? 30 seconds feels like it's on the high side to me. I'm general, the graph page processes a lot of data, and does so generally through lots of point queries to the database which isn't super fast.
The homepage took around this long to load. The 30s wait time in the screenshoot came from leaving the start and end dates empty and choosing "context-switches" as the desired metric.
Viewing graphs on perf.rust-lang.org often takes a long time, seemingly mainly spent waiting for a response from the server when loading data:
Are the current bottlenecks here already well understood?