I saw the response on http://www.spinics.net/lists/fio/msg04392.html (didn't get the email), so I'm pasting it here. I git pulled the latest fio code and observed the specified changes. I compiled and ran the latest code but still seeing the same problem: =============== test: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=1 fio-2.2.12-50-g5261 Starting 1 thread test: (groupid=0, jobs=1): err= 0: pid=10726: Tue Dec 8 13:40:21 2015 read : io=799740KB, bw=79966KB/s, iops=19991, runt= 10001msec slat (usec): min=0, max=11, avg= 0.10, stdev= 0.30 clat (usec): min=26, max=2624, avg=49.03, stdev=34.07 lat (usec): min=35, max=2624, avg=49.12, stdev=34.07 clat percentiles (usec): | 1.00th=[ 36], 5.00th=[ 36], 10.00th=[ 36], 20.00th=[ 37], | 30.00th=[ 37], 40.00th=[ 38], 50.00th=[ 38], 60.00th=[ 39], | 70.00th=[ 40], 80.00th=[ 70], 90.00th=[ 79], 95.00th=[ 87], | 99.00th=[ 92], 99.50th=[ 94], 99.90th=[ 225], 99.95th=[ 298], | 99.99th=[ 2128] bw (KB /s): min= 0, max=81072, per=95.01%, avg=75974.80, stdev=17896.56 lat (usec) : 50=71.32%, 100=28.42%, 250=0.19%, 500=0.04%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01% cpu : usr=99.99%, sys=0.00%, ctx=8, majf=0, minf=0 IO depths : 1=157.5%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=199935/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=1 Run status group 0 (all jobs): READ: io=799740KB, aggrb=79966KB/s, minb=79966KB/s, maxb=79966KB/s, mint=10001msec, maxt=10001msec =============== Jens Axboe wrote: On 11/25/2015 12:36 PM, Deyoung Hong (dhong) wrote: > I think this is an fio bug. > > Description: When specifying ramp_time, cpu may show over 100% (when using a polling mode driver). > If ramp_time is not specified, cpu values appear to be correct. > You normally don't see this problem because most drivers don't utilize high CPU usage. > You will have to use some custom driver that does polling, for example (something like http://www.dpdk.org/). > > OS: Any (but tested on Linux CentOS 6.6) This should fix it: http://git.kernel.dk/cgit/fio/commit/?id=210dd0fc66b8956065e098bb9314febe8993e849 Try current -git and hopefully it'll behave better. -- 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 -- 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
