Creating throughput graph along with ResponseTime graph

[JChav]

Hi Folks,

JChav is really helpful in getting the response time graphs after the
performance benchmarks.

It would be great if we can extend it to support throughput graph
generation also.

Another idea is to have median, standard deviation in the response time
graphs so that we can know the validity of the response time generated
for a specific benchmark.

Any thoughts on this folks?

~Rajesh.B

[dall...@gmail.com]

To take the second point first...

> Another idea is to have median, standard deviation in the response time
> graphs so that we can know the validity of the response time generated
> for a specific benchmark.

Yes, I agree we should be gathering this information. What we'd
probably need to do is find a chart that can represent all this
information in a clear way. Something like a box plot maybe... one to
think about.

> It would be great if we can extend it to support throughput graph
> generation also.

Sounds good! I'm not clear on how we generate a throughput graph: I
mean, I don't know what's involved in coming up with the numbers to
plot. If anyone has any pointers on this, that'd be much appreciated.

Richard

[Rajesh Balamohan]

Hi,

I believe we are generating the graph based on the output of the jmeter log (JTL file or some xml file which contains the response times etc).

TPS = Throughput per second. (ie, the number of requests that got executed in a second).

By default JMeter provides it in the "Aggregate report". They try to consolidate all the response times and divide it by the total time it took to execute the benchmark.

I hope we can come up with the similiar approach as well. We would be having the list of all response tmies for a specific benchmark. All we need to do is to sum it up and divide by the time taken to execute the benchmark.

This would be useful in the following case.

1. In performance world, people generally run benchmarks with multiple threads (1,2,4,8,12,16 threads etc).
2. After the runs are over, they plot the graph with the following information (X-axis - number of threads, Y-axis - Throughput per second)

This would give a clear picture on how the system is behaving under increasing load.

--
~Rajesh.B