Trying ot better understand when requested iodepth is not achieved

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

 



I recently started doing some higher iodepth (32/64) testing with some
high-end consumer SSDs. Of the 6 I tried, one consistently led to fio
runs achieving a lower percentage of IO at the requested iodepth.
(Every other SSD had >99% of IO at requested depth, This particular
drive would get as low as 97%.)

I was trying to understand how specific hardware could affect this. It
would seem like the only thing an SSD could do to cause lower
iodepth_level would be to process data faster than fio could supply
it. So I would expect times of unusually low latency.

Instead, looking at the runs when this happens, the drive is actually
performing unusually poorly. There are some big latency dips from high
latencies to much better latencies, but the latencies seen are never
anything lower than another better-performing drive.

Any thoughts? Happy to share graphs if helpful.

Thanks,
Nick




[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