From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from cantor2.suse.de ([195.135.220.15]:45922 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751007AbaEEQQM (ORCPT ); Mon, 5 May 2014 12:16:12 -0400 Date: Mon, 5 May 2014 18:16:11 +0200 From: David Sterba To: Hendrik Friedel Cc: linux-btrfs@vger.kernel.org Subject: Re: Very high load when reading/writing Message-ID: <20140505161611.GT5988@twin.jikos.cz> Reply-To: dsterba@suse.cz References: <53628953.8030707@friedels.name> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <53628953.8030707@friedels.name> Sender: linux-btrfs-owner@vger.kernel.org List-ID: On Thu, May 01, 2014 at 07:50:11PM +0200, Hendrik Friedel wrote: > I have very high load when writing/reading from/to two of my btrfs volumes. > One sda1, mounted as /mnt/BTRFS, the other, sdd2/sde2 (raid) as / > sda1 is a 3TB disc, whereas the sdd2/sde2 are small SSDs of 16GB. > > I wrote a small script to demonstrate it. It does: > -echo what it will do > -show the current load > -dd from one volume to the other. > -show the current load > -sync and flush the cache > -sleep 300s in order to get the load down again. load average is not a good metric, it mixes IO an CPU load into one number, would be better to measure cpu and IO loads separately. > Here the output: > Test from /mnt/BTRFS to /tmp > 124,553 s, 16,8 MB/s Is the transfer speed comparable if using a bare blockdevice instead of a filesystem on top of that? According to the S.M.A.R.T. stats the devices seem to be working normally.