Gehel added a comment.

Doing some quick math:

  • wdqs eqiad cluster has between 20 and 40 requests per seconds according to our varnish stats
  • 3 servers in that cluster
  • let's assume requests are routed equally to all servers, that's 10 req/second/server
  • 1.3 GB/sec allocation rate
  • 130MB allocated per request
  • the calculation above do not take writes into account, to the allocation per request is probably quite a bit lower
  • 130MB / request is high, but not entirely crazy

Note that I also find anecdotal evidence through Google that an allocation rate of 1GB/core is sustainable (this does not match my experience, where I've seen issues much below that).

I need to dig more into G1 documentation to see how to tune generations sizes to optimize for short lived objects (which seems to be our case).


TASK DETAIL
https://phabricator.wikimedia.org/T175919

EMAIL PREFERENCES
https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: Gehel
Cc: Stashbot, gerritbot, Smalyshev, Gehel, Aklapper, Lordiis, Lucas_Werkmeister_WMDE, GoranSMilovanovic, Adik2382, Th3d3v1ls, Ramalepe, Liugev6, QZanden, EBjune, merbst, Avner, Lewizho99, Maathavan, debt, Jonas, FloNight, Xmlizer, Izno, jkroll, Wikidata-bugs, Jdouglas, aude, Tobias1984, Manybubbles, Mbch331
_______________________________________________
Wikidata-bugs mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs

Reply via email to