From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: MIME-Version: 1.0 In-Reply-To: <6863f0c90906202031h54c9f74bwdb09ca084d05ddfa@mail.gmail.com> References: <7efa8a7d0906191519l5b268430o3ec4c96e60dfcb7a@mail.gmail.com> <20090620062702.GI31415@kernel.dk> <6863f0c90906200621l48000095xc46ef1fbcfc159d8@mail.gmail.com> <5048B694-088C-4C4A-B53A-95E250C004A5@kernel.dk> <7efa8a7d0906201338n7a70dc03y67cb956e5da5e765@mail.gmail.com> <7efa8a7d0906201417x4b3d3c76o91b52627da1e623c@mail.gmail.com> <6863f0c90906202031h54c9f74bwdb09ca084d05ddfa@mail.gmail.com> Date: Sat, 20 Jun 2009 22:15:56 -0700 Message-ID: <7efa8a7d0906202215s581f4f17t14afaac9f7aea2e6@mail.gmail.com> Subject: Re: Ques on FIO output (how to interpret iops) From: Learner Study Content-Type: multipart/alternative; boundary=0016e645ba423313b5046cd4da81 To: John Cagle Cc: Jens Axboe , fio@vger.kernel.org List-ID: --0016e645ba423313b5046cd4da81 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit I'm using 1.24 version - does latest 1.27 fix this issue? Thanks! On Sat, Jun 20, 2009 at 8:31 PM, John Cagle wrote: > Actually, 1KiB is 1024 bytes, but you're right -- the io values and the > aggrb values don't seem to agree. > > What version of fio are you using? There was a bug in one of the earlier > version (around 1.24 or 1.25) where it was using 1000 instead of 1024 when > doing some calculations. > > John > > > On Sat, Jun 20, 2009 at 4:17 PM, Learner Study wrote: > >> The summary says, io=1,723MiB, aggrb=30,105KiB/s >> >> So, aggrb = io/runtime => 1723/60 => 28.716 => 28, 716 KiB/s (assuming >> 1KiB = 1000 B)....still != aggrb shown in the result... >> >> Thanks! >> >> >> >> On Sat, Jun 20, 2009 at 1:38 PM, Learner Study wrote: >> >>> runtime = 60 and blocksize=512 >>> >>> The exact cmd is: >>> ./fio --name=global --runtime=60 --rw=read --bs=512 --direct=1 >>> --ioengine=aio --iodepth=4000 --loops=100 --name=job1 --filename=/dev/sdb >>> --name=job2 --filename=/dev/sdb >>> >>> Thanks! >>> >>> >>> >>> On Sat, Jun 20, 2009 at 7:12 AM, Jens Axboe wrote: >>> >>>> On 20/06/2009, at 15.21, John Cagle wrote: >>>> >>>> On Sat, Jun 20, 2009 at 1:27 AM, Jens Axboe < >>>> jens.axboe@oracle.com> wrote: >>>> >>>>> On Fri, Jun 19 2009, Learner Study wrote: >>>>> > Hi: >>>>> > >>>>> > How do I calculate total IOPS done by different FIO threads/jobs? >>>>> > >>>>> > Following is output from FIO (with two threads)... >>>>> > >>>>> > Each thread shows "bw" and "iops" numbers. Towards the end of output >>>>> log, >>>>> > there is "aggrb" which is sum of individual "bw" of threads, but "io" >>>>> field >>>>> > (3 line from the bottom of output below) doesn't seem to be sum of >>>>> > individual thread iops.... >>>>> > >>>>> > For example, Thread_1 has bw: 15k iops: 29366 >>>>> > Thread_2 has bw: 15k iops: 29445 >>>>> > But, the summary towards the end (third line from bottom of output), >>>>> says >>>>> > "io=1,723MiB, aggrb=30,105KiB/s". >>>>> > >>>>> > In this aggrb is good but io doesn't match up (it should be 58781) ! >>>>> Can >>>>> > someone please educate what am I missing? >>>>> >>>>> Can you try a newer version? If the problem is still there, the best >>>>> way >>>>> to get it fixed is to send my a simple job file that reproduces the >>>>> problem, and then I can easily fix it for you. >>>>> >>>> >>>> I thought the "io=" summary at the end was the total amount of I/O >>>> performed (in megabytes), not the total number of IOPS. >>>> >>>> >>>> It is, you are correct. Seems I misread the problem. >>>> >>> >>> >> > --0016e645ba423313b5046cd4da81 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable I'm using 1.24 version - does latest 1.27 fix this issue?

Thanks= !

On Sat, Jun 20, 2009 at 8:31 PM, John C= agle <jcagle@gmail= .com> wrote:

Actually, 1KiB is= 1024 bytes, but you're right -- the io values and the aggrb values don= 't seem to agree.

What version of fio are you using?=A0 There was a bug in one of the ear= lier version (around 1.24 or 1.25) where it was using 1000 instead of 1024 = when doing some calculations.

John

On Sat, Jun 20, 2009 at 4:17 PM, Learner Study &= lt;learner.stu= dy@gmail.com> wrote:

The summary says, io=3D1,723MiB, aggrb=3D30,105KiB/s

So, aggrb =3D i= o/runtime =3D> 1723/60 =3D> 28.716 =3D> 28, 716 KiB/s (assuming 1K= iB =3D 1000 B)....still !=3D aggrb shown in the result...

Thanks!

On Sat, Jun 20, 2009 at 1:38 PM, Learner Study <learner.study@gmail.= com> wrote:
runtime =3D 60 and blocksize=3D512

The exact cmd is:
./fio --name=3Dglobal --runtime=3D60 --rw=3Dread --bs=3D512 --direct=3D1 --ioengine=3Daio --iodepth=3D4000 --loops=3D100 --name=3Djob1 --filename=3D/dev/sdb --name=3Djob2 --filename=3D/dev/sdb

Thanks!

On Sat, Jun 20, 2009 at 7:12 AM, Jens Ax= boe <axboe@kernel.dk> wrote:

On 20/06/2009, at = 15.21, John Cagle <jcagle@gmail.com> wrote:

On Sat, Jun 20, 2009 at 1:27 AM, Jens = Axboe <= jens.axboe@oracle.com> wrote:

On Fri, Jun 19 2009, Learner Study wrote:
> Hi:
>
> How do I calculate total IOPS done by different FIO threads/jobs?
>
> Following is output from FIO (with two threads)...
>
> Each thread shows "bw" and "iops" numbers. Towards= the end of output log,
> there is "aggrb" which is sum of individual "bw" o= f threads, but "io" field
> (3 line from the bottom of output below) doesn't seem to be sum of=
> individual thread iops....
>
> For example, Thread_1 has bw: 15k iops: 29366
> Thread_2 has bw: 15k iops: 29445
> But, the summary towards the end (third line from bottom of output), s= ays
> "io=3D1,723MiB, aggrb=3D30,105KiB/s".
>
> In this aggrb is good but io doesn't match up (it should be 58781)= ! Can
> someone please educate what am I missing?

Can you try a newer version? If the problem is still there, the best way to get it fixed is to send my a simple job file that reproduces the
problem, and then I can easily fix it for you.

I = thought the "io=3D" summary at the end was the total amount of I/= O performed (in megabytes), not the total number of IOPS.

It is, you are correct. Seems I misread the prob= lem.

--0016e645ba423313b5046cd4da81--