public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Bernardo Innocenti <bernie@develer.com>
To: lkml <linux-kernel@vger.kernel.org>, OpenLDAP-devel@OpenLDAP.org
Cc: Giovanni Bajo <rasky@develer.com>, Simone Zinanni <simone@develer.com>
Subject: sched_yield() makes OpenLDAP slow
Date: Thu, 18 Aug 2005 02:50:16 +0200	[thread overview]
Message-ID: <4303DB48.8010902@develer.com> (raw)

Hello,

I've been investigating a performance problem on a
server using OpenLDAP 2.2.26 for nss resolution and
running kernel 2.6.12.

When a CPU bound process such as GCC is running in the
background (even at nice 10), many trivial commands such
as "su" or "groups" become extremely slow and take a few
seconds to complete.

strace revealed that data exchange over the slapd socket
was where most of the time was spent.  Looking at the
slapd side, I see several calls to sched_yield() like this:


[pid  8780]      0.000033 stat64("gidNumber.dbb", 0xb7b3ebcc) = -1 EACCES (Permission denied)
[pid  8780]      0.000059 pread(20, "\0\0\0\0\1\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\2\0\344\17\2\3"..., 4096, 4096) = 4096
[pid  8780]      0.000083 pread(20, "\0\0\0\0\1\0\0\0\4\0\0\0\3\0\0\0\0\0\0\0\222\0<\7\1\5\370"..., 4096, 16384) = 4096
[pid  8780]      0.000078 time(NULL)    = 1124322520
[pid  8780]      0.000066 pread(11, "\0\0\0\0\1\0\0\0\250\0\0\0\231\0\0\0\235\0\0\0\16\0000"..., 4096, 688128) = 4096
[pid  8780]      0.000241 write(19, "0e\2\1\3d`\4$cn=bernie,ou=group,dc=d"..., 103) = 103
[pid  8780]      0.000137 sched_yield( <unfinished ...>
[pid  8781]      0.050020 <... sched_yield resumed> ) = 0
[pid  8780]      0.000025 <... sched_yield resumed> ) = 0
[pid  8781]      0.000060 futex(0x925ab20, FUTEX_WAIT, 33, NULL <unfinished ...>
[pid  8780]      0.000026 write(19, "0\f\2\1\3e\7\n\1\0\4\0\4\0", 14) = 14
[pid  8774]      0.000774 <... select resumed> ) = 1 (in [19])


The relative timestamp reveals that slapd is spending 50ms
after yielding.  Meanwhile, GCC is probably being scheduled
for a whole quantum.

Reading the man-page of sched_yield() it seems this isn't
the correct behavior:

   Note: If the current process is the only process in the
   highest priority list at that time, this process will
   continue to run after a call to sched_yield.

I also think OpenLDAP is wrong.  First, it should be calling
pthread_yield() because slapd is a multithreading process
and it just wants to run the other threads.  See:

int
ldap_pvt_thread_yield( void )
{
#if HAVE_THR_YIELD
        return thr_yield();

#elif HAVE_PTHREADS == 10
        return sched_yield();

#elif defined(_POSIX_THREAD_IS_GNU_PTH)
        sched_yield();
        return 0;

#elif HAVE_PTHREADS == 6
        pthread_yield(NULL);
        return 0;
#else
        pthread_yield();
        return 0;
#endif
}


-- 
  // Bernardo Innocenti - Develer S.r.l., R&D dept.
\X/  http://www.develer.com/


             reply	other threads:[~2005-08-18  0:50 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-08-18  0:50 Bernardo Innocenti [this message]
2005-08-18  0:47 ` sched_yield() makes OpenLDAP slow Con Kolivas
2005-08-18 10:48   ` Maciej Soltysiak
2005-08-18  1:07 ` Joseph Fannin
2005-08-18  2:25   ` Bernardo Innocenti
2005-08-18  2:58   ` Nick Piggin
2005-08-19  3:10     ` Bernardo Innocenti
     [not found] <4303DB48.8010902@develer.com.suse.lists.linux.kernel>
     [not found] ` <20050818010703.GA13127@nineveh.rivenstone.net.suse.lists.linux.kernel>
     [not found]   ` <4303F967.6000404@yahoo.com.au.suse.lists.linux.kernel>
     [not found]     ` <43054D9A.7090509@develer.com.suse.lists.linux.kernel>
2005-08-19  3:19       ` Andi Kleen
2005-08-19  3:30         ` Bernardo Innocenti
  -- strict thread matches above, loose matches on Subject: below --
2005-08-19  6:03 Howard Chu
2005-08-19  6:34 ` Nick Piggin
2005-08-19  6:59 ` Chris Wedgwood
2005-08-19 22:45   ` Howard Chu
2005-08-19 10:21 ` Nikita Danilov
2005-08-19 23:10   ` Howard Chu
2005-08-20 13:23     ` Nikita Danilov
2005-08-20 19:49       ` Howard Chu
2005-08-20 22:08         ` Nikita Danilov
     [not found] <4D8eT-4rg-31@gated-at.bofh.it>
2005-08-20  3:20 ` Robert Hancock
2005-08-20  4:18   ` Nick Piggin
2005-08-20 18:38     ` Howard Chu
2005-08-20 20:57       ` Lee Revell
2005-08-20 21:24         ` Howard Chu
2005-08-21  0:36           ` Nick Piggin
2005-08-21 11:33           ` Nikita Danilov
2005-08-22  8:06             ` Howard Chu
2005-08-20 21:50       ` Lee Revell
2005-08-21  1:04       ` Robert Hancock
2005-08-22 11:44         ` linux-os (Dick Johnson)
2005-08-22 14:26           ` Robert Hancock
2005-08-23 11:17             ` linux-os (Dick Johnson)
2005-08-23 12:07               ` Denis Vlasenko
2005-08-24  3:37                 ` Lincoln Dale
     [not found] <43057641.70700@symas.com.suse.lists.linux.kernel>
     [not found] ` <17157.45712.877795.437505@gargle.gargle.HOWL.suse.lists.linux.kernel>
     [not found]   ` <430666DB.70802@symas.com.suse.lists.linux.kernel>
2005-08-20 13:48     ` Andi Kleen
2005-08-21 19:47       ` Florian Weimer
2005-08-22  5:09         ` Howard Chu
2005-08-22 13:06           ` Andi Kleen
2005-08-22 18:47             ` Howard Chu
2005-08-22 13:20           ` Florian Weimer
2005-08-22 23:19             ` Howard Chu
2006-01-14 19:29 Howard Chu
     [not found] <5uZqb-4fo-15@gated-at.bofh.it>
2006-01-14 22:47 ` Robert Hancock
2006-01-24 22:59 e100 oops on resume Stefan Seyfried
2006-01-24 23:21 ` Mattia Dongili
2006-01-25  9:02   ` Olaf Kirch
2006-01-25 12:11     ` Olaf Kirch
2006-01-25 13:51       ` sched_yield() makes OpenLDAP slow Howard Chu
2006-01-25 14:38         ` Robert Hancock
2006-01-25 17:49         ` Christopher Friesen
2006-01-26  1:07         ` David Schwartz
2006-01-26  8:30           ` Helge Hafting
2006-01-26  9:01             ` Nick Piggin
2006-01-26 10:50             ` Nikita Danilov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4303DB48.8010902@develer.com \
    --to=bernie@develer.com \
    --cc=OpenLDAP-devel@OpenLDAP.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=rasky@develer.com \
    --cc=simone@develer.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox