mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Bill Rugolsky Jr." <brugolsky@telemetry-investments.com>
To: Jakob Oestergaard <jakob@unthought.net>,
	Trond Myklebust <trond.myklebust@fys.uio.no>,
	linux-kernel@vger.kernel.org
Subject: Re: NFS client (10x) performance regression 2.6.14.7 -> 2.6.15
Date: Fri, 31 Mar 2006 11:35:42 -0500	[thread overview]
Message-ID: <20060331163542.GC9207@ti64.telemetry-investments.com> (raw)
In-Reply-To: <20060331152401.GM9811@unthought.net>

On Fri, Mar 31, 2006 at 05:24:01PM +0200, Jakob Oestergaard wrote:
> With a local export/mount I see very few reads running the test  :/
> 
> Could this be a timing issue?  That pages in the cache are invalidated
> after maybe less than 100 microseconds which is roughly the round-trip
> time on the gigabit network connecting my server and client?

I don't see any problem over loopback either.  Perhaps I'll try increasing
delays with tc netem.  That will have to wait until later.

In any case, I tried on our NetApp FAS-250, and I see the same problem:
Mount options are:
netapp:/home /nfs/netapp/home nfs rw,v3,rsize=32768,wsize=32768,hard,intr,tcp,lock,addr=netapp 0 0

I don't have stat zeroing, so I rolled some scripts.   Sorry, these are
not vanilla kernels; if you don't nail this today, I'll try and find time
for git bisection over the weekend.

	-Bill

== 

My production FC1 2.6.10 kernel:

ti64: rm -f testfile ; uname -a ; ( /tmp/nfsstat.sh ; time /tmp/nfsbench ; /tmp/nfsstat.sh ) | /tmp/nfsstatdiff.awk 
Linux ti64 2.6.10-0.ti.4.fc1smp #1 SMP Wed Jan 19 12:15:34 EST 2005 i686 athlon i386 GNU/Linux

real    0m0.215s
user    0m0.000s
sys     0m0.093s
null: 0 getattr: 2 setattr: 0 lookup: 1 access: 0 readlink: 0 read: 1 write: 321 create: 1 mkdir: 0 symlink: 0 mknod: 0 remove: 0 rmdir: 0 rename: 0 link: 0 readdir: 0 readdirplus: 0 fsstat: 0 fsinfo: 0 pathconf: 0 commit: 0 

===

DaveJ's FC4 2.6.15.3-based errata kernel, mounted:

ti63: rm -f testfile ; uname -a ; ( /tmp/nfsstat.sh ; time /tmp/nfsbench ; /tmp/nfsstat.sh ) | /tmp/nfsstatdiff.awk 
Linux ti63 2.6.15-1.1831_FC4smp #1 SMP Tue Feb 7 13:48:31 EST 2006 i686 athlon i386 GNU/Linux

real    0m3.242s
user    0m0.000s
sys     0m0.360s
null: 0 getattr: 2561 setattr: 0 lookup: 1 access: 1 readlink: 0 read: 7040 write: 3200 create: 1 mkdir: 0 symlink: 0 mknod: 0 remove: 0 rmdir: 0 rename: 0 link: 0 readdir: 0 readdirplus: 0 fsstat: 0 fsinfo: 0 pathconf: 0 commit: 0 


	-Bill

  reply	other threads:[~2006-03-31 16:35 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-03-31  9:48 Jakob Oestergaard
2006-03-31 11:22 ` NFS client regression, simple test program Jakob Oestergaard
2006-03-31 12:22 ` NFS client (10x) performance regression 2.6.14.7 -> 2.6.15 Trond Myklebust
2006-03-31 12:45   ` Jakob Oestergaard
2006-03-31 13:06     ` Trond Myklebust
2006-03-31 13:21       ` Jakob Oestergaard
2006-03-31 13:44         ` Trond Myklebust
2006-03-31 14:08           ` Jakob Oestergaard
2006-03-31 14:21             ` Trond Myklebust
2006-03-31 14:35               ` Jakob Oestergaard
2006-03-31 14:49                 ` Bill Rugolsky Jr.
2006-03-31 14:57                   ` Jakob Oestergaard
2006-03-31 15:04                     ` Bill Rugolsky Jr.
2006-03-31 15:24                       ` Jakob Oestergaard
2006-03-31 16:35                         ` Bill Rugolsky Jr. [this message]
2006-03-31 18:52                           ` Jakob Oestergaard
2006-03-31 15:55                 ` Trond Myklebust
2006-03-31 16:04                   ` Jakob Oestergaard
2006-04-03 15:26                     ` Jakob Oestergaard
2006-04-03 15:41                       ` Trond Myklebust
2006-04-03 15:45                         ` Jakob Oestergaard
2006-04-04  9:22                           ` Jakob Oestergaard
2006-04-24 22:03                             ` Trond Myklebust
2006-04-25  8:08                               ` Jakob Oestergaard

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=20060331163542.GC9207@ti64.telemetry-investments.com \
    --to=brugolsky@telemetry-investments.com \
    --cc=jakob@unthought.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=trond.myklebust@fys.uio.no \
    /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

all inboxes | Powered by JetHome®