From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail190.messagelabs.com (mail190.messagelabs.com [216.82.249.51]) by kanga.kvack.org (Postfix) with SMTP id CB3866B0047 for ; Wed, 24 Feb 2010 06:18:28 -0500 (EST) Received: by ywh33 with SMTP id 33so5894313ywh.11 for ; Wed, 24 Feb 2010 03:18:27 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <20100224052215.GH16175@discord.disaster> References: <20100224024100.GA17048@localhost> <20100224032934.GF16175@discord.disaster> <20100224041822.GB27459@localhost> <20100224052215.GH16175@discord.disaster> Date: Wed, 24 Feb 2010 06:18:26 -0500 Message-ID: Subject: Re: [RFC] nfs: use 2*rsize readahead size From: Akshat Aranya Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Sender: owner-linux-mm@kvack.org To: Dave Chinner Cc: Wu Fengguang , Trond Myklebust , "linux-nfs@vger.kernel.org" , "linux-fsdevel@vger.kernel.org" , Linux Memory Management List , LKML List-ID: On Wed, Feb 24, 2010 at 12:22 AM, Dave Chinner wrote: > >> It sounds silly to have >> >> =A0 =A0 =A0 =A0 client_readahead_size > server_readahead_size > > I don't think it is =A0- the client readahead has to take into account > the network latency as well as the server latency. e.g. a network > with a high bandwidth but high latency is going to need much more > client side readahead than a high bandwidth, low latency network to > get the same throughput. Hence it is not uncommon to see larger > readahead windows on network clients than for local disk access. > > Also, the NFS server may not even be able to detect sequential IO > patterns because of the combined access patterns from the clients, > and so the only effective readahead might be what the clients > issue.... > In my experiments, I have observed that the server-side readahead shuts off rather quickly even with a single client because the client readahead causes multiple pending read RPCs on the server which are then serviced in random order and the pattern observed by the underlying file system is non-sequential. In our file system, we had to override what the VFS thought was a random workload and continue to do readahead anyway. Cheers, Akshat -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org