To: Trond Myklebust <trond.myklebust@fys.uio.no>
cc: linux-kernel@vger.kernel.org
Subject: Re: 2.6.0 NFS-server low to 0 performance
From: Andi Kleen <ak@muc.de>
Date: Sat, 10 Jan 2004 17:08:41 +0100
In-Reply-To: <1csrv-Er-9@gated-at.bofh.it> (Trond Myklebust's message of
 "Sat, 10 Jan 2004 15:40:09 +0100")
Message-ID: <m3eku74whi.fsf@averell.firstfloor.org>
User-Agent: Gnus/5.090013 (Oort Gnus v0.13) Emacs/21.2 (i586-suse-linux)
References: <1cpDr-5az-11@gated-at.bofh.it> <1csrv-Er-9@gated-at.bofh.it>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 768
Lines: 19

Trond Myklebust <trond.myklebust@fys.uio.no> writes:

> The correct solution to this problem is (b). I.e. we convert mount to
> use TCP as the default if it is available. That is consistent with what
> all other modern implementations do.

Please do that. Fragmented UDP with 16bit ipid is just russian roulette at 
today's network speeds.

One disadvantage is that some older (early 2.4) Linux nfsd servers that have
TCP enabled can cause problems. But I guess we can live with that, they
should be updated anyways.

-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/