Hi, I am trying to verify the statistic in the output log from the latency and iops logs write_lat_log write_iops_log when I import the iops data and did the statistic - I have a perfect match from FIO out log: iops : min= 37, max= 138, avg=123.63, stdev=11.31, samples=2668 from statistical tool calculation: Descriptive Statistics: iops Variable N N* CumN CumPct Mean SE Mean StDev Minimum Q1 iops 2668 0 2668 100 123.63 0.219 11.31 37.00 122.00 Variable Median Q3 Maximum iops 125.00 130.00 138.00 But when I tried to do the same exercise using the lat, slat and clat logs - the results are very different from FIO out log: lat (msec): min=3, max=442, avg= 7.98, stdev= 6.15 from statistical tool calculation: Descriptive Statistics: lat (ns) Variable N N* CumN CumPct Mean SE Mean StDev Minimum lat (ns) 2669 0 2669 100 31675586 1192150 61589285 8428382 Variable Q1 Median Q3 Maximum lat (ns) 15552716 18330303 24137899 442001336 May I know if there is filtering of data (outlier, etc) for that? My main objective is to understand the long tail latency (99.95% and 99.99%). The latency log has the timestamp and latency data and I just need to find out where the long latencies are coming from - beginning, middle or end of the test. Thanks in advance, TK
