heph
heph copied to clipboard
Thundering herd problem in shared timers
Pr #376 (knowingly) introduced the "Thundering herd problem" for the shared timers. If multiple worker threads have the same timeout (based on the shared timers) and all attempt to schedule the same thread-safe actor because of it will pointlessly waste resources, as the actor can only be scheduled once.
This should be resolved when fixing #284.
Fixed in #423.
Reopened by commit e8c0e158c838f04e1b402e7e3f904520e724d819.