[1180] in linux-net channel archive
Re: NFS blocksize
daemon@ATHENA.MIT.EDU (Richard Gooch)
Mon Oct 9 17:29:46 1995
Date: Mon, 9 Oct 1995 19:01:49 +1000
From: Richard Gooch <rgooch@atnf.CSIRO.AU>
To: iialan@iifeak.swan.ac.uk (Alan Cox)
Cc: linux-net@vger.rutgers.edu
In-Reply-To: <m0t2Dlj-000141C@iiit.swan.ac.uk>
Alan Cox writes:
> > Hi, all. I can't remember whether or not the old problem with NFS
> > transfers with buffersize != 1k still exists. Currently I mount NFS
> > filesystems with the default options (buffersize = 1k) and the
> > performance is *quite* poor. Can someone tell me if this was fixed?
>
> As of about 1.1.3 kmalloc() was fixed to allow big allocations, as of
> about 1.1.50 or so NFS large blocks works right. I use 8K to sunos
> without problems except when using a 3c501 where it consistently misses
> some of the fragments each time.
So why isn't 8k blocks the default now?
Regards,
Richard....
