Understanding of response time in Gatling

In Gatling if I want to collect response time for all data points is it possible to get that ? Percentile thing should be post aggregation after completion of test. Every second /unit sending 50th/75th/nth percentile doesn’t make any sense to me.is there anything I am missing

As I remember there should be global response times with percentiles.

As a side point I noticed a company using Gatling as a PaaS and providing average response times. That's a real problem if there are numerous high outliers. Best described in a scatter graph, I suppose.

I would be offering my stakeholders p99.+ as a key metric. That is 99% plus of users experiencing response time at this amount.

Many Thanks


Through graphite protocol Gatling is sending some percentile data points in timeseries database so I am not getting what is the meaning of percentile during run itself.these metrics should be calculated after run completion.

The goal of this Graphite integration is to provide a quick way to have live feedback. It has lots of drawbacks, be it because of the limitations of the Graphite wire protocol or the capabilities of the target database timeseries (lack of histogram objects in InfluxDB for example).

We have all those features and concerns covered in Gatling Enterprise: live reporting, storage, correct distributions computations.

You'd also want to be looking at Java Heap, file descriptor use, I/O, memory and CPU use etc, and ideally I'd want those metrics in one place.