Subject: Re: NFS/RPC and server clusters
To: Matthias Drochner <M.Drochner@fz-juelich.de>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: tech-net
Date: 10/15/2003 12:31:39
On Wed, Oct 15, 2003 at 06:19:49PM +0200, Matthias Drochner wrote:
> 
> zakhrin@freeshell.org said:
> > Have you tried to use NFS over TCP instead of UDP? This way the client
> > would stablish the connection to the service IP. 
> 
> I had thought of it, and now I've tried and it works indeed.
> (The kernel has do do a reconnect on the TCP socket after the
> failover which is a little bit more noisy.)
> 
> But it should work with UDP as well, because there is no point
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^
> in paying the TCP overhead in a local network and because it
  ^^^^^^^^^^^^^^^^^^^^^^^^^^

Actually, there is.  If you're using large RPC requests, TCP will
segment them neatly into multiple underlying IP packets; UDP will
generate gigantic packets that must then be fragmented and reassembled.

Aside from the inherent inefficiency, this also defeats checksum offload
on most network adapters that can do checksum offload.  Oops!

-- 
 Thor Lancelot Simon	                                      tls@rek.tjls.com
   But as he knew no bad language, he had called him all the names of common
 objects that he could think of, and had screamed: "You lamp!  You towel!  You
 plate!" and so on.              --Sigmund Freud