Subject: Re: nfs performance
To: Werner Backes <email@example.com>
From: Arto Huusko <firstname.lastname@example.org>
Date: 02/10/2004 19:12:50
Werner Backes wrote:
> Arto Huusko wrote:
>> Now, finally fed up with this, I tweaked with my nfs client
>> parameters. I now mount all my NFS volumes with a read size of
>> 2048 bytes, which seems to work snappily enough. 4096, and things
>> won't work anymore. Oh, and I've been using UDP all the time.
> Are you using NFS V2 or V3? I've always achived best performance for
> copying large files by using NFS3 over TCP with it's default blocksize
> of 32k.
> NFS2 is slower especially for writes. Small blocksizes (<8k) also make
> slower, but there may be circumstances where this is needed.
I'm using V3 over UDP. I also tried V3 over TCP with 8k readsize,
and it failed just as miserably.
Once I found that 2k works over UDP, I stopped trying anything
else. And it seems to be fast enough.
And in any case anybodys interested, the machines are on the same
subnet and there's only a dump switch between them.