From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:55910) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1YkUw1-0003sB-Sz for qemu-devel@nongnu.org; Tue, 21 Apr 2015 05:57:42 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1YkUvy-0005hG-Ol for qemu-devel@nongnu.org; Tue, 21 Apr 2015 05:57:41 -0400 Received: from mx1.redhat.com ([209.132.183.28]:49752) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1YkUvy-0005h6-JO for qemu-devel@nongnu.org; Tue, 21 Apr 2015 05:57:38 -0400 Message-ID: <55361F0D.9040907@redhat.com> Date: Tue, 21 Apr 2015 11:57:33 +0200 From: Paolo Bonzini MIME-Version: 1.0 References: <552E1EB4.3030805@clodo.ru> <20150416012735.GB21291@ad.nay.redhat.com> <552F9A4F.90504@clodo.ru> <552FA39C.7080209@redhat.com> <55361E1C.7040302@clodo.ru> In-Reply-To: <55361E1C.7040302@clodo.ru> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] virtio-blk and virtio-scsi performance comparison List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: kkv@clodo.ru, Fam Zheng Cc: qemu-devel@nongnu.org On 21/04/2015 11:53, Konstantin Krotov wrote: >> Can you detail the problem more? That could be a kernel bug. > > I connect to the host system, two storage systems as block devices. Both > block devices combine md-raid1. md-device use as disk for virtual > machine. If i use io=native, read perform only from one of two raid1 > legs. I'd like to distribute the load reading between two raid1 legs, > but md-raid1 implementation it does only if set io=thread. That looks like a kernel problem. Paolo