Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754184Ab0BBTib (ORCPT ); Tue, 2 Feb 2010 14:38:31 -0500 Received: from 0122700014.0.fullrate.dk ([95.166.99.235]:40040 "EHLO kernel.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751797Ab0BBTia (ORCPT ); Tue, 2 Feb 2010 14:38:30 -0500 Date: Tue, 2 Feb 2010 20:38:26 +0100 From: Jens Axboe To: Wu Fengguang Cc: Andrew Morton , Peter Zijlstra , Linux Memory Management List , linux-fsdevel@vger.kernel.org, LKML Subject: Re: [PATCH 01/11] readahead: limit readahead size for small devices Message-ID: <20100202193826.GC5733@kernel.dk> References: <20100202152835.683907822@intel.com> <20100202153316.375570078@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20100202153316.375570078@intel.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2982 Lines: 79 On Tue, Feb 02 2010, Wu Fengguang wrote: > Linus reports a _really_ small & slow (505kB, 15kB/s) USB device, > on which blkid runs unpleasantly slow. He manages to optimize the blkid > reads down to 1kB+16kB, but still kernel read-ahead turns it into 48kB. > > lseek 0, read 1024 => readahead 4 pages (start of file) > lseek 1536, read 16384 => readahead 8 pages (page contiguous) > > The readahead heuristics involved here are reasonable ones in general. > So it's good to fix blkid with fadvise(RANDOM), as Linus already did. > > For the kernel part, Linus suggests: > So maybe we could be less aggressive about read-ahead when the size of > the device is small? Turning a 16kB read into a 64kB one is a big deal, > when it's about 15% of the whole device! > > This looks reasonable: smaller device tend to be slower (USB sticks as > well as micro/mobile/old hard disks). > > Given that the non-rotational attribute is not always reported, we can > take disk size as a max readahead size hint. We use a formula that > generates the following concrete limits: > > disk size readahead size > (scale by 4) (scale by 2) > 2M 4k > 8M 8k > 32M 16k > 128M 32k > 512M 64k > 2G 128k > 8G 256k > 32G 512k > 128G 1024k I'm not sure the size part makes a ton of sense. You can have really fast small devices, and large slow devices. One real world example are the Sun FMod SSD devices, which are only 22GB in size but are faster than the Intel X25-E SLC disks. What makes it even worse for these devices is that they are often attached to fatter controllers than ahci, where command overhead is larger. Running your script on such a device yields (I enlarged the read-count by 2, makes it more reproducible): MARVELL SD88SA02 MP1F rasize 1st 2nd ------------------------------------------------------------------ 4k 41 MB/s 41 MB/s 16k 85 MB/s 81 MB/s 32k 102 MB/s 109 MB/s 64k 125 MB/s 144 MB/s 128k 183 MB/s 185 MB/s 256k 216 MB/s 216 MB/s 512k 216 MB/s 236 MB/s 1024k 251 MB/s 252 MB/s 2M 258 MB/s 258 MB/s 4M 266 MB/s 266 MB/s 8M 266 MB/s 266 MB/s So for that device, 1M-2M looks like the sweet spot, with even needing 4-8M to fully reach full throughput. I don't think this is atypical of bigger systems. Only very recently have controller started to slim down the command overhead for real, because of the SSD devices. What probably is atypical is a device that is this small yet pretty fast. -- Jens Axboe -- 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/