NFS server - /proc/net/rpc/nfsd

[Asaf Maruf]

Hi
Here are some tips you could use to improve overall performance:

16 is a good number to start with. Increasing the number of threads for
non-Solaris OS can create performance issues as memory is allocated upfront
and threads are activated on every request.

Run nfsstat -rc on the client. The value under TimeOuts should be small
indicating a healthy network.  Ensure that both server and clients are
running the current NFS version 3 or 4. Can also force server and clients to
use TCP like so:
mount -o proto=tcp server:/export    /nfs

 Using the mount command, the read and write blocks can be increased to 32k
for better performance

Some recommended values for improving NFS performance:

MTU size increase to 4000-9000 bytes
NFS block size 8KBytes
Increasing hard disk speed by using hdparm


Hope this is helpful.
Asaf


On Feb 19, 2008 1:02 PM, Rob Sutherland <sutherland_rob-/E1597aS9LQAvxtiuMwx3w at public.gmane.org> wrote:

> I'm trying to nail down an NFS problem that seems to
> be related to load and one of the recommendations in
> the NFS Howto is to increase the number of NFS server
> instances, which are currently at the default, 8.
> Anyone have a rule of thumb for the number of
> instances?
>
> The Howto also recommends looking at the statistics in
> the /proc/net/rpc/nfsd file. I'm not really sure how
> to interperet these numbers, can anyone point me to an
> explanation?
>
> Thanks,
>
> Rob
>
>
>
> Rob Sutherland
>
> 'I is Popeye O'Borg.
>  Yez will be askimigulated'
>
>
>
>  ____________________________________________________________________________________
> Be a better friend, newshound, and
> know-it-all with Yahoo! Mobile.  Try it now.
> http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
>
> --
> The Toronto Linux Users Group.      Meetings: http://gtalug.org/
> TLUG requests: Linux topics, No HTML, wrap text below 80 columns
> How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists
>



-- 
"I can live with doubt and uncertainty and not knowing. I think it is much
more interesting to live not knowing than to have answers that might be
wrong." - Richard P. Feynman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gtalug.org/pipermail/legacy/attachments/20080219/44ade212/attachment.html>