On Mon, Jul 23 2007, Erwan Velu wrote:
> Jens Axboe wrote:
>> Can't reproduce. If I were to venture a guess, I would say that it's a
>> combination of:
>>
>> - small bw window (it's only 500 msec)
>> - larger block size (64kb)
>> - other activity on the drive?
>>
>> Was the tested drive otherwise idle?
>>
> yes it was totally dedicated to fio but the test was quite long.
> I did a test on a raid controller, my device was 1.16 TB big.
It would be interesting to see if you could see a dip in vmstat output
as well, if you repeat the test!
> Maybe fio failed on such bug array.
Size should not matter, it's just an offset question and fio should be
able to handle a full 64-bit offset just fine.
-- Jens AxboeReceived on Mon Jul 23 2007 - 15:11:50 CEST
This archive was generated by hypermail 2.2.0 : Mon Jul 23 2007 - 15:30:02 CEST