From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from aserp1040.oracle.com ([141.146.126.69]:19749 "EHLO aserp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751010AbcHLVRn (ORCPT ); Fri, 12 Aug 2016 17:17:43 -0400 To: One Thousand Gnomes Cc: "Martin K. Petersen" , Tom Yan , Tejun Heo , jmoyer@redhat.com, axboe@fb.com, linux-ide@vger.kernel.org, linux-scsi@vger.kernel.org, linux-block@vger.kernel.org, Sergei Shtylyov Subject: Re: [PATCH v2 2/2] libata-core: do not set dev->max_sectors for LBA48 devices From: "Martin K. Petersen" References: <5eceeb9ad825beb87db03b0989f746dfbe0cf3fe.1470753817.git.tom.ty89@gmail.com> <10c4d38005e24bb477b5182d4fc17879139f14d9.1470753817.git.tom.ty89@gmail.com> <20160810041042.GH25053@mtj.duckdns.org> <20160812101645.0a4a5e9f@lxorguk.ukuu.org.uk> Date: Fri, 12 Aug 2016 17:17:22 -0400 In-Reply-To: <20160812101645.0a4a5e9f@lxorguk.ukuu.org.uk> (One Thousand Gnomes's message of "Fri, 12 Aug 2016 10:16:45 +0100") Message-ID: MIME-Version: 1.0 Content-Type: text/plain Sender: linux-block-owner@vger.kernel.org List-Id: linux-block@vger.kernel.org >>>>> "Alan" == One Thousand Gnomes writes: Alan, Alan> For older SCSI and especially ATA drives (and it wouldn't surprise Alan> me if it is true of modern ones) there are also huge latency Alan> tradeoffs. Absolutely. Alan> Before you jump up and down about numbers what are the latency Alan> numbers like on classic ATA drives with that sized block I/O. You Alan> could easily up your RAID numbers while wrecking realtime and Alan> desktop performance. I was in no way advocating raising the default, quite the contrary. We have several customer workloads that got negatively impacted when the limit was bumped. My point was that arguing about what the default should be without any data supporting the discussion is futile. Alan> At the ATA level we can detect both the presence of command Alan> queueing ability, and also whether the device is spinnning rust or Alan> not, so it may be smarter defaults could be done based upon Alan> whether the device is an SSD or not. Sure. However, modern SSDs actually trend towards smaller I/O sizes. NVMe is 128K by default, I think. So a higher default wouldn't make any difference there. -- Martin K. Petersen Oracle Linux Engineering