Re: [Question] What if the parameter for --size option is smaller than the target device?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On 21 September 2014 04:57, Akira Hayakawa <ruby.wktk@xxxxxxxxx> wrote:
> I am writing a test with fio tool (2.0.8) and I am facing a problem.
>
> I just want to submit 512B random write to a block device.
> In my case, the size of the device is 2MB and the amount of writes (specified by --size option) is 32MB.

You probably wanted the behaviour of io_limit
(http://git.kernel.dk/?p=fio.git;a=blob;f=HOWTO;h=73e58ff695c3384df30094d6c15ebe9ed615460f;hb=HEAD#l433
) rather than size...

> like this (A line from device-mapper-test-suite. Is my command wrong?):
>       ProcessControl.run("fio --name=test --filename=#{dev.path} --rw=randwrite --ioengine=libaio --direct=1 --size=#{@param[0]}m --bs=512")
>
> The problem is, the runtime is too short.
> The block device is just a single HDD so, it will be 30sec or so. But it's actually 3sec...
>
> My guess is that fio uses min(device size, param to --size) as the amount of writes here.

That's correct in this context - size is limiting the last possible
block that can be written to (see
http://git.kernel.dk/?p=fio.git;a=blob;f=HOWTO;h=73e58ff695c3384df30094d6c15ebe9ed615460f;hb=HEAD#l421
for details).

> To investigate it, I increased the size of the device to x10 and saw the runtime gets x10.
>
> My questions are:
> 1. Is my guess correct?

Yes.

> 2. What shall I do to make fio work as I want it to do?

Use io_limit if your fio is new enough or number_ios
(http://git.kernel.dk/?p=fio.git;a=blob;f=HOWTO;h=73e58ff695c3384df30094d6c15ebe9ed615460f;hb=HEAD#l778
).

Sitsofe | http://sucs.org/~sits/
--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Kernel]     [Linux SCSI]     [Linux IDE]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux SCSI]

  Powered by Linux