NFS Tuning

[James Cooley]

Brian,

I've got a couple of ideas.  The first one, is to try to improve your 
send and receive packet sizes.  This usually defaults to 4K, but 8K 
tends to have a dramatic improvement on performance, and users commonly 
set it using the rsize and wsize options like so when mounting from the 
client-side:

mount -t nfs -o rsize=8192,wsize=8192    ...

If you have a more modern version of Linux at both ends, you can try 
increasing the sizes even further.


Another possibility is to increase the size of the NFS memory buffers on 
the server to say 256KB.  On RHEL 4, the default is 128KB.

You can do this by using the following commands:

   echo 262144 > /proc/sys/net/core/rmem_default
   echo 262144 > /proc/sys/net/core/rmem_max



Thanks,
James Cooley



Brian D. McGrew wrote:

>Good morning,
>
> 
>
>Can someone help me with tuning up NFS or point me to a good reference?
>I'm running several servers, RH7.3, RH9, Solaris 8 and FC3.  We're
>getting absolutely terrible network performance and it's not just with
>NFS.  However, I ran up ethereal on a few of the servers and in less
>then three minutes I captured over a million packets, of which 95.4%
>were UDP NFS packets.  How do I go about turning up NFS?  I know this is
>way too much traffic for the size of our network and I'm at a complete
>loss as to what to do.
>
> 
>
>Thanks,
>
> 
>
>-brian
>
> 
>
>Brian D. McGrew { brian at visionpro.com || brian at doubledimension.com }
>
>---
>
>  
>
>>YOU!  Off my planet!
>>    
>>
>
> 
>
>  
>


-- 
--
James Cooley
Sr. Systems Analyst
Information Technology
Florida Tech
321-674-7999
jcooley at it.fit.edu