From: Neil Brown Subject: Re: [PATCH][RFC] Multiple UDP sockets for knfsd Date: Sun, 16 May 2004 08:24:31 +1000 Sender: nfs-admin@lists.sourceforge.net Message-ID: <16550.39071.745015.438121@cse.unsw.edu.au> References: <40A42DB8.7B9C8152@melbourne.sgi.com> <1084566110.4237.66.camel@lade.trondhjem.org> <20040515063854.GB14983@sgi.com> <1084641723.3490.13.camel@lade.trondhjem.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Greg Banks , Linux NFS Mailing List Return-path: Received: from sc8-sf-mx1-b.sourceforge.net ([10.3.1.11] helo=sc8-sf-mx1.sourceforge.net) by sc8-sf-list2.sourceforge.net with esmtp (Exim 4.30) id 1BP7aY-0000KQ-ED for nfs@lists.sourceforge.net; Sat, 15 May 2004 15:25:02 -0700 Received: from note.orchestra.cse.unsw.edu.au ([129.94.242.24] ident=root) by sc8-sf-mx1.sourceforge.net with esmtp (Exim 4.30) id 1BP7aX-0001aU-WE for nfs@lists.sourceforge.net; Sat, 15 May 2004 15:25:02 -0700 Received: From notabene.cse.unsw.edu.au ([129.94.241.7] == ppp2-007.ppp2.cse.unsw.EDU.AU) (auth-user neilb) (for ) (for ) (for ) By note With Smtp ; Sun, 16 May 2004 08:24:47 +1000 To: Trond Myklebust In-Reply-To: message from Trond Myklebust on Saturday May 15 Errors-To: nfs-admin@lists.sourceforge.net List-Unsubscribe: , List-Id: Discussion of NFS under Linux development, interoperability, and testing. List-Post: List-Help: List-Subscribe: , List-Archive: On Saturday May 15, trond.myklebust@fys.uio.no wrote: > On Sat, 2004-05-15 at 02:38, Greg Banks wrote: > > On Fri, May 14, 2004 at 04:21:50PM -0400, Trond Myklebust wrote: > > > > > > I'm confused here. Are you actually implying that socket contention is > > > responsible for a full *40%* decrease in throughput? > > > > At that level of throughput, yes. > > So what exactly *is* this bottleneck? As far as we're concerned, a > socket is basically just a buffer and a couple of locks. > It is almost certainly svsk->sk_sem. We hold this while calling sock->ops->sendpage on each individual page of the reply - we don't want them mixed up with some other reply. Remember that the test here involves lots of large READ requests. Each will have a few pages (this is an ia64 with large pages, so there will probably be only one or two pages or page-fragments of data) what will need to be fragmented into multiple UDP packets. NeilBrown ------------------------------------------------------- This SF.Net email is sponsored by: SourceForge.net Broadband Sign-up now for SourceForge Broadband and get the fastest 6.0/768 connection for only $19.95/mo for the first 3 months! http://ads.osdn.com/?ad_id=2562&alloc_id=6184&op=click _______________________________________________ NFS maillist - NFS@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs