Abhishek, For a good explanation of how fio histograms work: https://github.com/axboe/fio/blob/master/stat.h#L30 I really like the approach that allows you to trade off accuracy and latency range, where there is a potentially enormous latency range. Bucket latency range doubles for successive bucket groups. For buckets in each bucket group, latency range is fixed. There is a compile-time limit on the number of buckets at present. you can recompile fio and change FIO_IO_U_PLAT_GROUP_NR. The limit of 16 seconds comes from this compile time limit. Would be nice if row 1 of the histogram log contained the upper bound of each histogram bucket's latency range. You can parse the histogram logs and generate latency percentiles as a function of time from them with fiologparser_hist.py. HTH -ben ----- Original Message ----- > From: "abhishek koundal" <akoundal@xxxxxxxxx> > To: fio@xxxxxxxxxxxxxxx > Sent: Tuesday, March 7, 2017 1:49:17 AM > Subject: fio output-format=json+ and latency histogram buckets ~1200 > > Hi, > Can someone please explain what is the detailed latency buckets using > --output-format=json+ depict? The are always ~16 buckets populated. I > am not able to figure out the # iops per latency buckets and what the > units of latency are? > > Another question: > For the histogram latency there are ~1200 buckets which can track > 16sec for max latency. Can someone please explain what each bucket > interval. e.g. assuming each is 10us increment. > 0-10us [#IOPS] > 10-20us[#IOPS] > 20-30us..[#IOPS] > ...... > 15-16sec > -- > Life is too short for silly things so invest your time in some > productive outputs!! > -- > 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
