[Beowulf] diskless cluster nfs

[Ed Hill]

On Tue, 2004-12-07 at 17:23 -0500, Josh Kayse wrote:
> I've recently been tasked to build a diskless cluster for one of our
> engineers.  This was easy because we already had an image for the set
> of machines.  Once we started testing, the performance was very poor. 
> Basic setup follows:

Hi Josh,

We've spent some time tuning NFS on our clusters and still have NFS
performance thats somewhere between "not so good" and "awful" relative
to raw disk speeds.  The short list of things to check is:

 - async option (which you're using--good)
 - increasing the number of nfsd processes
 - having an SMP (or even HyperThreaded) NFS server

All other things equal, we'd seen substantially better NFS performance
(throughput) when using SMP NFS servers.

Also, you may want to look into lustre if throughput is a big concern.

Ed

-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS;  Rm 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464