From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.viktor.guru ([62.210.97.36]:14260 "EHLO mail.viktor.guru" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751514AbbJONLl (ORCPT ); Thu, 15 Oct 2015 09:11:41 -0400 Received: from localhost (localhost [127.0.0.1]) by mail.viktor.guru (Postfix) with ESMTP id 887E812A095B for ; Thu, 15 Oct 2015 15:11:39 +0200 (CEST) Received: from mail.viktor.guru ([127.0.0.1]) by localhost (mail.viktor.guru [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id W1LsXsuJo804 for ; Thu, 15 Oct 2015 15:11:38 +0200 (CEST) Received: from [10.10.0.21] (unknown [91.185.100.46]) by mail.viktor.guru (Postfix) with ESMTPSA id 6303412A0951 for ; Thu, 15 Oct 2015 15:11:37 +0200 (CEST) Subject: FIO blocksize option clear out References: <561F97E4.5030807@htec.rs> From: vmokric Message-ID: <561FA608.3010904@htec.rs> Date: Thu, 15 Oct 2015 15:11:36 +0200 MIME-Version: 1.0 In-Reply-To: <561F97E4.5030807@htec.rs> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Sender: fio-owner@vger.kernel.org List-Id: fio@vger.kernel.org To: fio@vger.kernel.org Hello members of the fio forum. In fio man it says that: blocksize=int[,int], bs=int[,int] Block size for I/O units. Default: 4k. Values for reads and writes can be specified separately in the format/read/,/write/, either of which may be empty to leave that value at its default. What exactly is blocksize, i.e. block size for I/O units? I'm rather new in this world of data storages and file systems, I/O schedulers, block layers, etc., and I'm rather confused with the block size terminology. Judging by wiki page on block devices and block sizes: http://wiki.linuxquestions.org/wiki/Block_devices_and_block_sizes, what FIO impacts over its block size parameter is File system block size, right? I saw that people also give another name for file system block size: "cluster size". So, by default block size (File system block size) is 4k. And my disk physical sector size (disk block size) is also 4k (and there is another thing: Kernel block size, and I really don't know how this impacts data transfer to and from my sda disk (with or without caching) or I have some vague idea...). Let's say that I change my fio block size to 1m. My disk block size is still 4k. What is really the difference here now, what is changed? Can someone explain or point me to some source? I know it's rather newbie question... And, does IO depth depends on my SATA disk physical characteristics? Can I go higher that IO depth = 32 if my disk returns: root@linaro-gnome:~# hdparm -I /dev/sda | grep Queue Queue depth: 32 * Native Command Queueing (NCQ) Thanks, Vojislav