Mean/Average response time as charting metric


I am currently doing a performance test for a new version of the application which is being hosted on the AWS cloud. The previous benchmark tests that I have access to have the stats reported in Avg response time at any instance ( tool used was Visual Studio Load Test - Visual Studio enables Avg response time by default but has option to turn on 95th and 99th percentile too (disabled by default) ).

While I run the gatling tests, the charting (to a graphite server) has options of 95th, 99th, min and max response time. Is it possible to add Average/Mean to chart (I see that there is no such option on the gatling.conf file)? If not, what would be a good way for me to compare my results (in 95th percentile) and the previous results (in average) to deduce if the application performance has improved or degraded on any particular ec2 instance?



IIRC, we finally ditched average value, because we thought this metric, though widely used, is actually not very meaningful: a good average doesn’t mean that you don’t have have your users/requests experience very bad performance while the other half is fine. IMHO, percentiles are a much better indicator of your quality of service.

Discussion is open, of course. :slight_smile:

Note that you do get the average in the standard reports.



Second that. .