I'm testing ceveral configurations of a server. At each run there is a script that is making hunderds of curl request. After running the tests per each configuration I have the following output:

  1. Number of requests
  2. Number os success
  3. Time of each request. I'm calculating the avarege and median (using median because there is a very high variance or outlier as suggestsed in another thread)
  4. CPU consumption of the server.

How can I make a simple graph that will show the results for my managers (non technichal)? More specificaly, I can calculate the avarege/median throuput of each configuration but how decide which is better when taking into account the number of success? As an example, configuration A completed 100 files downloads in an avarege of 1 second and configuration B completed 130 files in an avarege of 1.2 seconds. Which is better? How to display it?

If someone can suggest a way to put inside the servers' CPU consumption it will be a real winner. Thanks!


I'll try to answer your 2 questions here,

How to get stats like mean/median of requests per config and plot charts?

For this, Load testing tools like JMeter, Gatling, LoadRunner with graph support are better. Their advantages as compared to curl scripts are ,

  1. They are more efficient/sophisticated in terms of understanding, sharing and are extensible, scalable. In short its standard practice in load testing.
  2. These tools support various other protocols
  3. Scenarios like caching, failure handling, response validation are handled out of the box.
  4. Stats calculation and graph generation is available as plugins or out of box in most of the above.

How to get resource utilization of servers

For this system monitoring tools are used. Examples are ganglia, munin, graphite-influxdb, sar etc. They have agents installed on cluster nodes and will send resource util data periodically to monitoring server. Plotting can be done using Grafana.

  • $\begingroup$ I thought that this answers my question but I need to make my own calculations. So thanks for the info but there is a reason why I asked it here as a statistical question and in tools/utilities forum. $\endgroup$ – nmnir Apr 18 '18 at 10:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.