From: Bernd Schubert <bernd-schubert@web.de>
Subject: async vs. sync
Date: Mon, 26 Jul 2004 17:17:59 +0200
Sender: nfs-admin@lists.sourceforge.net
Message-ID: <200407261718.03264.bernd-schubert@web.de>
References: <200407231620.i6NGKYV20945@algebra.hillsboro.credence.com>
Mime-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
To: Linux NFS Mailing List <nfs@lists.sourceforge.net>
In-Reply-To: <200407231620.i6NGKYV20945@algebra.hillsboro.credence.com>
Errors-To: nfs-admin@lists.sourceforge.net


> On a completely seperate note, Linux-to-Linux NFS file writes
> only seem to be fast if we publish the serving Linux volume
> as asynchronous (default setting is synchronous which is slow).
>

Yeah, we just observed something similar.=20

Here some numbers for write speed of the clients:

linux-2.6.7:
async: 11MB/s
sync, wdelay: 2-3MB/s
sync, no_wdelay: 7MB/s

linux-2.4.27-rc3:
async: 11MB/s
sync, wdelay: 2-3MB/s
sync, no_wdelay: 2-3MB/s

The kernel version corresponds to the server kernel, all clients still run=
=20
2.4.26.

Unfortunality switching the server to 2.6.7 makes the server crash every=20
morning on running the cron-jobs with page allocation errors, so keeping=20
2.6.x is currently not an option. At least not as long as those errors seem=
=20
to be tolerated by the kernel maintainers :(

Any ideas whats the issue with sync vs async in 2.4.x?

Thanks,
	Bernd


=2D-=20
Bernd Schubert
Physikalisch Chemisches Institut / Theoretische Chemie
Universit=E4t Heidelberg
INF 229
69120 Heidelberg
e-mail: bernd.schubert@pci.uni-heidelberg.de


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs