From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@bugzilla.kernel.org Subject: [Bug 12752] megaraid-sas driver performance worse than on-board sata_nv Date: Fri, 10 Jun 2011 23:05:04 GMT Message-ID: <201106102305.p5AN54hl028552@demeter2.kernel.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Return-path: Received: from demeter2.kernel.org ([140.211.167.42]:56606 "EHLO demeter2.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758214Ab1FJXFF (ORCPT ); Fri, 10 Jun 2011 19:05:05 -0400 Received: from demeter2.kernel.org (localhost.localdomain [127.0.0.1]) by demeter2.kernel.org (8.14.4/8.14.4) with ESMTP id p5AN54Pv028553 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Fri, 10 Jun 2011 23:05:04 GMT In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: linux-scsi@vger.kernel.org https://bugzilla.kernel.org/show_bug.cgi?id=12752 Adam Radford changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |aradford@gmail.com --- Comment #2 from Adam Radford 2011-06-10 23:05:02 --- devsk, I'm sorry about you not getting a response from the previous megaraid_sas developer. He was probably not watching bugzilla.kernel.org. There is nothing in the driver that will 'throttle' the I/O on your PERC5 controller. The software raid code sets various read-ahead values in the block layer request_queues it sets up: # grep -i "ra_pages" * raid0.c: * If an individual device has an ra_pages greater than the raid0.c: if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) raid0.c: mddev->queue->backing_dev_info.ra_pages = 2* stripe; raid10.c: if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) raid10.c: mddev->queue->backing_dev_info.ra_pages = 2* stripe; raid5.c: if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe) raid5.c: mddev->queue->backing_dev_info.ra_pages = 2 * stripe; raid5.c: if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe) raid5.c: conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe; Can you try setting some OS read-ahead values (on top of the controller doing read-ahead) for the scsi device attached to your PERC5 controller and re-running your tests? i.e.: /sbin/blockdev --getra /dev/sda (See what you have set currently) /sbin/blockdev --setra 8192 /dev/sda (Set read ahead for sda to 8192 sectors) /sbin/blockdev --setra 16834 /dev/sda (Set read ahead for sda to 16384 sectors) This may cause more I/O to be issued to the card from the OS and may improve your bonnie++ results. -Adam -- Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching the assignee of the bug.