linux-cifs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Fwd: Deferred close patch - performance improvements
       [not found] <CAH2r5mtjCnXosTKnV7J33=T4aVxD4kkkU6PcuHisMw=trLycbg@mail.gmail.com>
@ 2021-04-16 18:55 ` Steve French
  0 siblings, 0 replies; only message in thread
From: Steve French @ 2021-04-16 18:55 UTC (permalink / raw)
  To: CIFS

---------- Forwarded message ---------
From: Steve French <smfrench@gmail.com>
Date: Fri, Apr 16, 2021 at 1:55 PM
Subject: Deferred close patch - performance improvements
To: rohiths msft <rohiths.msft@gmail.com>


I noticed some significant improvements in perf of some xfstest cases
on the latest buildbot run (and some test cases where perf got worse,
but fewer cases of that) with Rohith's recent deferred close patch (it
is possible, although less likely that some other patches in for-next
contributed to the performance changes - it is probably mostly due to
Rohith's patch). Some examples:

cifs/104 improved from 28 to 4 seconds, 85% faster
generic/011 from 8:24 to 7:28, 11% faster
generic/032 from 8:10 to 5:34, 32% faster
generic/117 from 5:35 to 5:01, 10% faster
generic/239 from 1:41 to 1:22 19% faster

But a few got much slower:
generic/070 from 5:08 to 8:48
generic/100 from 3:44  to 5:10
generic/263 from 3:32 to 10:17

Although there is some variable in perf on individual tests due to the
usual temporary load factors, overall the run was 10 to 20 minutes
faster (5 to 10% faster) than usual (which will be helpful going
forward in multiple ways) with current for-next, and is promising.
--
Thanks,

Steve


-- 
Thanks,

Steve

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2021-04-16 18:55 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <CAH2r5mtjCnXosTKnV7J33=T4aVxD4kkkU6PcuHisMw=trLycbg@mail.gmail.com>
2021-04-16 18:55 ` Fwd: Deferred close patch - performance improvements Steve French

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).