Conversation
Notices
-
Ruby Rhod (feld@bikeshed.party)'s status on Sunday, 04-Aug-2019 00:27:02 UTC
Ruby Rhod
@sjw @lanodan yeah but that sucks a lot.
Your single core is being shared between a ton of processes and the L1/2/3 are never going to have cache hits, especially if being shared between other VPSes. Also BEAM intentionally takes more CPU time than needed for soft real-time so there is going to be Pleroma work wasting CPU time doing nothing and the kernel isn't even switching to Postgres to process the query.