I am running a load test for 30 minutes. Once the load test is completed, i get the analysis file.
1. We generally provide the 95th Percentile for the Response Time for the transaction clicks.
2. From server details, mean of the respective counter is used.
My question is how can we compare the 95th percentile of the RT as with mean of the server counters monitored.?
From my understanding but not sure though i think in both the cases it should be 95th percentile. Kindly let me know your ideas too.
What is sampling rate of CPU? Since CPU spikes highs and lows and could be extreme if your sampling rate is low (but good for such low durtion test to identify bottelneck)ůso percentile does not make much sense...
Ramesh, do you want to cut off the (5%) of the worst CPU usage stats? You could theoretically do that, but I never do that because as A$h mentioned you are anyway doing sampling resource counters (average over period of time), while response times is a set actual numbers. There is no sense to do percentile over averages, is there?
?:the art of a constructive conflict perceived as a destructive diagnose.