On 2014-09-28 15:45, Hinson, Roger wrote:
Hey all, I'm setting up a test to simulate what we're actually seeing on one of our storage arrays. Can you take a look and provide me some feedback? The array usage is for two workload types: Type 1: Random Read % - 45 Random Write % - 55 Sequential Read % - 1 Sequential Write % - 3 R/W Ratio - 45:55 Random Block Size - 4K Sequential Block Size - 64K Type 2: Random Read % - 1 Random Write % - 13 Sequential Read % - 2 Sequential Write % - 84 R/W Ratio - 3:97 Random Block Size - 4K Sequential Block Size - 32K I came up with this: [global] group_reporting=1 thread iodepth=8 ioengine=libaio direct=1 refill_buffers randrepeat=1 randseed=100 random_distribution=zipf:1.01 runtime=1200 time_based bs_is_seq_rand [45read55write_dm0_numjobs_2] write_bw_log=4555_dm0_2 write_lat_log=4555_dm0_2 write_iops_log=4555_dm0_2 filename=/dev/mapper/mpathq rw=randrw rwmixread=45 rwmixwrite=55 percentage_random=96 bs=64K,4K buffer_compress_percentage=50 buffer_compress_chunk=512 numjobs=2 [3read97write_dm1_numjobs_2] write_bw_log=397_dm1_2 write_lat_log=397_dm1_2 write_iops_log=397_dm1_2 filename=/dev/mapper/mpathr rw=randrw rwmixread=3 rwmixwrite=97 percentage_random=14 bs=32K,4K buffer_compress_percentage=50 buffer_compress_chunk=512 numjobs=2
This looks like a pretty good approximation. Note that percentage_random can take all data directions into account. So if you have a workload that is:
read sequential: 57% write sequential: 95% you could do: percentage_random=43,5 and have that just work. -- Jens Axboe -- 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
