From: Chris Penney <cpenney@gmail.com>
To: nfs@lists.sourceforge.net
Subject: NFS Client Performance Question
Date: Wed, 13 Jul 2005 14:30:00 -0400 [thread overview]
Message-ID: <111aefd050713113071f0c9ce@mail.gmail.com> (raw)
I have a large smp SGI Itanium box box running a 2.4.21 kernel and I'm
getting fairly bad NFS performance, esp. when doing random writes to
an unloaded NFS server. The box is rather busy doing computation and
a lot of i/o to /tmp; however, there is little network i/o (<100
KB/s). I use iozone with the following options to test "-c -e -i 2 -w
-s 16m" and use an unloaded Sun NFS server (same performance results
with Linux NFS servers, but they are all under load so I test with the
Sun). I only get ~2.5MB/s using the above test. I found an Intel box
still running a 2.4.20 kernel and it gets 15MB/s. On a 2.6 kernel box
the rate jumps to 46MB/s (nice work).
I did some sniffing on the Sun and noticed that the problem Itanium
system was not doing async writes (snips are from the start of the
write test):
[snip]
redhat -> server NFS C WRITE3 FH=3D40E5 at 1478656 for 4096 (ASYNC)
server -> redhat NFS R WRITE3 OK 4096 (ASYNC)
redhat -> server NFS C WRITE3 FH=3D40E5 at 1179648 for 4096 (ASYNC)
[snip]
itanium -> server NFS C WRITE3 FH=3D40E5 at 4239360 for 4096 (FSYNC)
server -> server NFS R WRITE3 OK 4096 (FSYNC)
server -> server NFS C WRITE3 FH=3D40E5 at 15101952 for 4096 (FSYNC)
[snip]
The client mount options are:
rw,nosuid,bg,hard,intr,nfsvers=3D3,tcp,rsize=3D32768,wsize=3D32768
So my question: Is the Itanium system not doing async nfs i/o because
nfract_sync has been exceeded due to local i/o to /tmp or is it
something else? Is there anything I can do to improve NFS performance
on this box?
Chris
-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA. To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
NFS maillist - NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs
next reply other threads:[~2005-07-13 18:30 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-07-13 18:30 Chris Penney [this message]
2005-07-13 20:16 ` NFS Client Performance Question Trond Myklebust
2005-07-13 20:40 ` Chris Penney
-- strict thread matches above, loose matches on Subject: below --
2005-07-13 18:44 Lever, Charles
2005-07-13 18:55 ` Chris Penney
2005-07-13 19:09 ` Peter Staubach
2005-07-13 19:57 ` Chris Penney
2005-07-13 19:04 Lever, Charles
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=111aefd050713113071f0c9ce@mail.gmail.com \
--to=cpenney@gmail.com \
--cc=nfs@lists.sourceforge.net \
--cc=penney@msu.edu \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.