From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Subject: Re: file_service_type and random_distribution References: <5739FC0D.7070106@kernel.dk> <573A0D4D.8010205@kernel.dk> From: Jens Axboe Message-ID: <573A71E4.3020404@kernel.dk> Date: Mon, 16 May 2016 19:20:36 -0600 MIME-Version: 1.0 In-Reply-To: <573A0D4D.8010205@kernel.dk> Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit To: qingwei wei , fio@vger.kernel.org List-ID: On 05/16/2016 12:11 PM, Jens Axboe wrote: > On 05/16/2016 10:57 AM, Jens Axboe wrote: >>> I would like to confirm those points is because i intend to run >>> workload test that simulate hot file access. At any one time, the >>> files open will be less than the total files in the system and there >>> will be files that are hot and get access multiple times. Can i do it >>> using fio? Thanks. >> >> That's a good question. Currently fio doesn't, but there's no reason why >> it could not use the same distribution methods for choosing what files >> to service. That would indeed allow you to mimic some files being hotter >> than others, just like we do for offsets in that file. > > Something like the below patch would allow you to do that. Basically the > options and values are idential to random_distribution, this one is just > for selecting what file to service. Just like with random_distribution, > you could use fio-genzipf to give you a good idea of what the > distribution of file uses would be. > > Let me know how it works for you. Committed with some fixes, so you can just use the git version. -- Jens Axboe