Offset in percentage rather than in LBA

Offset in percentage rather than in LBA

[Prabhakaran]

Hi,

Can offset be specified in percentage rather than in capacity value
like gb or mb?
I have a config file which fills a drive once randomly. I got 4 jobs
running in parallel (4*64qd=256qd) with lba space divided among them.
I want to specify in percentage, so that all the jobs are equally
used, rather than hard-coding the capacity.

The following example configuration is designed for 450GB NVMe drive.


#################################################
[global]
ioengine=libaio
thread
cpus_allowed=0-7
cpus_allowed_policy=split
direct=1
size=100g
loops=6

[4k-4k_randwrite_0_rd_qd256_numjob1]
bs=4k
ba=4k
iodepth=64
offset=0
rw=randwrite
randrepeat=0
norandommap
refill_buffers
log_avg_msec=1000
unified_rw_reporting=1
write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob1
write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob1

[4k-4k_randwrite_0_rd_qd256_numjob2]
bs=4k
ba=4k
iodepth=64
offset=100g
rw=randwrite
randrepeat=0
norandommap
refill_buffers
log_avg_msec=1000
unified_rw_reporting=1
write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob2
write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob2

[4k-4k_randwrite_0_rd_qd256_numjob3]
bs=4k
ba=4k
iodepth=64
offset=200g
rw=randwrite
randrepeat=0
norandommap
refill_buffers
log_avg_msec=1000
unified_rw_reporting=1
write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob3
write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob3

[4k-4k_randwrite_0_rd_qd256_numjob4]
bs=4k
ba=4k
iodepth=64
offset=300g
rw=randwrite
randrepeat=0
norandommap
refill_buffers
log_avg_msec=1000
unified_rw_reporting=1
write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob4
write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob4
#################################################


Thanks,
Prabhu

Re: Offset in percentage rather than in LBA

[Sitsofe Wheeler]

On 27 January 2017 at 21:58, Prabhakaran <prabhugce@gmail.com> wrote:
>
> Can offset be specified in percentage rather than in capacity value
> like gb or mb?

Not at the moment (fio-2.17) but being able to do this was also
requested over in https://github.com/axboe/fio/issues/89 .

-- 
Sitsofe | http://sucs.org/~sits/

Re: Offset in percentage rather than in LBA

[Prabhakaran]

Thanks Tim!! I think, it's the best alternate way at this point and it works :)

On 27 January 2017 at 15:50, Tim Walker <tim.t.walker@seagate.com> wrote:
> I can't answer your question directly, but one could call fio from a BASH
> script that sets environment variables appropriately. Here is an example of
> dividing an ATA drive into 4 regions and setting environment variables to
> the starts. You can add offset=${offset_1} to your basic fio script then run
> it with these environmental variable set. My script assumes 512 byte
> sectors.
>
> (I didn't actually try the fio part, but I did test the BASH portion, so
> YMMV)
>
> bytes_per_sector=512
> maxlba=`hdparm -g /dev/sda | grep sectors | awk '{print $6}' | tr ',' ' '`
>
> quarter=$((${maxlba} / 4 * ${bytes_per_sector}))
> offset_1=0
> offset_2=$(($quarter * 1))
> offset_3=$(($quarter * 2))
> offset_4=$(($quarter * 3))
>
> echo $offset_1
> echo $offset_2
> echo $offset_3
> echo $offset_4
>
>
>
> On Fri, Jan 27, 2017 at 2:58 PM, Prabhakaran <prabhugce@gmail.com> wrote:
>>
>> Hi,
>>
>> Can offset be specified in percentage rather than in capacity value
>> like gb or mb?
>> I have a config file which fills a drive once randomly. I got 4 jobs
>> running in parallel (4*64qd=256qd) with lba space divided among them.
>> I want to specify in percentage, so that all the jobs are equally
>> used, rather than hard-coding the capacity.
>>
>> The following example configuration is designed for 450GB NVMe drive.
>>
>>
>> #################################################
>> [global]
>> ioengine=libaio
>> thread
>> cpus_allowed=0-7
>> cpus_allowed_policy=split
>> direct=1
>> size=100g
>> loops=6
>>
>> [4k-4k_randwrite_0_rd_qd256_numjob1]
>> bs=4k
>> ba=4k
>> iodepth=64
>> offset=0
>> rw=randwrite
>> randrepeat=0
>> norandommap
>> refill_buffers
>> log_avg_msec=1000
>> unified_rw_reporting=1
>> write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob1
>> write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob1
>>
>> [4k-4k_randwrite_0_rd_qd256_numjob2]
>> bs=4k
>> ba=4k
>> iodepth=64
>> offset=100g
>> rw=randwrite
>> randrepeat=0
>> norandommap
>> refill_buffers
>> log_avg_msec=1000
>> unified_rw_reporting=1
>> write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob2
>> write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob2
>>
>> [4k-4k_randwrite_0_rd_qd256_numjob3]
>> bs=4k
>> ba=4k
>> iodepth=64
>> offset=200g
>> rw=randwrite
>> randrepeat=0
>> norandommap
>> refill_buffers
>> log_avg_msec=1000
>> unified_rw_reporting=1
>> write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob3
>> write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob3
>>
>> [4k-4k_randwrite_0_rd_qd256_numjob4]
>> bs=4k
>> ba=4k
>> iodepth=64
>> offset=300g
>> rw=randwrite
>> randrepeat=0
>> norandommap
>> refill_buffers
>> log_avg_msec=1000
>> unified_rw_reporting=1
>> write_bw_log=4k-4k_randwrite_0_rd_qd256_numjob4
>> write_iops_log=4k-4k_randwrite_0_rd_qd256_numjob4
>> #################################################
>>
>>
>> Thanks,
>> Prabhu
>> --
>> To unsubscribe from this list: send the line "unsubscribe fio" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at
>> https://urldefense.proofpoint.com/v2/url?u=http-3A__vger.kernel.org_majordomo-2Dinfo.html&d=DwIBaQ&c=IGDlg0lD0b-nebmJJ0Kp8A&r=NW1X0yRHNNEluZ8sOGXBxCbQJZPWcIkPT0Uy3ynVsFU&m=0gTEIsl_-d0HQ0ndt8ua9fiIR5Kigo0mnZKBy3FrqS0&s=NZyL5aNn-frWDTJuCfS7XH7peqTcbHBIGubxUH6A0DM&e=
>
>
>
>
> --
> Tim Walker
> Product Design Systems Engineering, Seagate Technology
>



-- 
Thanks,
Prabhakaran