From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:39651) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1R4A6u-0004t1-FV for qemu-devel@nongnu.org; Thu, 15 Sep 2011 07:28:08 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1R4A6q-0001EP-Ev for qemu-devel@nongnu.org; Thu, 15 Sep 2011 07:28:04 -0400 Received: from mail-vw0-f45.google.com ([209.85.212.45]:51099) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1R4A6q-0001EL-BB for qemu-devel@nongnu.org; Thu, 15 Sep 2011 07:28:00 -0400 Received: by vws17 with SMTP id 17so4479292vws.4 for ; Thu, 15 Sep 2011 04:27:59 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <4E71C146.6070406@redhat.com> References: <1315628610-28222-1-git-send-email-ronniesahlberg@gmail.com> <1315628610-28222-2-git-send-email-ronniesahlberg@gmail.com> <20110912091408.GA3465@stefanha-thinkpad.localdomain> <4E71956B.6030902@redhat.com> <4E71BBE3.5000207@redhat.com> <4E71C146.6070406@redhat.com> Date: Thu, 15 Sep 2011 21:27:59 +1000 Message-ID: From: ronnie sahlberg Content-Type: text/plain; charset=ISO-8859-1 Subject: Re: [Qemu-devel] [PATCH] This patch adds a new block driver : iSCSI List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini Cc: kwolf@redhat.com, Stefan Hajnoczi , dlaor@redhat.com, qemu-devel@nongnu.org, fujita.tomonori@lab.ntt.co.jp, Orit Wasserman , hch@lst.de On Thu, Sep 15, 2011 at 7:11 PM, Paolo Bonzini wrote: ... > Perhaps Ronnie has rough performance numbers comparing in-kernel iSCSI with > libiscsi? I did some tests some time ago just doing things like 'dd if=/dev/sda of=/dev/null' on the root disk from within a RHEL guest itself and performace was comparable. In some tests libiscsi was a few percent faster, in other tests it was a few percent slower. But overall, very unscientifically performed test, performance was not much different. Now, I am not an expert on tuning cache or tuning the kernel scsi and iscsi layer, so I would not really want to make too many statements in the area other than in some tests I performed, performance was good enough for my use. I would be very happy if someone competent in tuning qemu/scsi/open-iscsi (==that means, not me) would do a real comparasion. regards ronnie sahlberg