Hi, I don't know why the minimum clat is 0 usec (you might be able to investigate further by using blktrace on the device to see how much that disagrees with fio) but why do you believe that scenario happened on the exact same I/O(s) that had a 0 usec slat time? Do you have additional logs showing that was the case? On 24 March 2017 at 17:39, Tarek <cashew250@xxxxxxxxx> wrote: > Clat min is zero on a random read, how is that possible when NAND tR > spec is 20-60uS? Also isn't lat = slat + clat, 1 + 0 != 84 > > fio-2.15 > Starting 80 threads > > random-100r-0w_4k_qd1_jobs_16: (groupid=0, jobs=16): err= 0: > pid=22233: Tue Feb 21 15:41:33 2017 > Description : [Random R/W 100/0 QD1 BLK 4k Jobs 16] > read : io=1045.5GB, bw=609017KB/s, iops=152254, runt=1800001msec > slat (usec): min=1, max=6583, avg= 2.06, stdev= 1.01 > clat (usec): min=0, max=8728, avg=102.38, stdev=21.43 > lat (usec): min=84, max=8730, avg=104.52, stdev=21.45 > clat percentiles (usec): > | 1.00th=[ 85], 5.00th=[ 86], 10.00th=[ 87], 20.00th=[ 89], > | 30.00th=[ 91], 40.00th=[ 98], 50.00th=[ 101], 60.00th=[ 102], > | 70.00th=[ 104], 80.00th=[ 106], 90.00th=[ 116], 95.00th=[ 143], > | 99.00th=[ 175], 99.50th=[ 185], 99.90th=[ 225], 99.95th=[ 239], > | 99.99th=[ 270] > > > Fio Config: > thread > bs=4k > ioengine=libaio > direct=1 > buffered=0 > log_avg_msec=1000 > group_reporting=1 > rw=randread > numjobs=16 > iodepth=1 > time_based > runtime=1800 > refill_buffers > norandommap > filename=/dev/nvme0n1 -- 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
