1
votes

I am trying to perform a load test in one of our applications.

The test is really simple with 1 thread group with 3 HTTP requests.

  • Login (After login, wait for 5 seconds)
  • Order (After Order, wait for 5 seconds)
  • Logout (After logout, wait for 5 seconds)
  • Repeat the above 3 steps again and again for the entire test duration.

I tried running the test for various users (100 users for 2 hrs, another test for 200 users for 2 hrs, another test for 300 users for 3 hours).

My throughput / response time graph always has huge fluctuation in the results.

enter image description here

I am not seeing issues specific to any of the above 3 HTTP requests. Their response time are all within 150 millsec to 70 seconds. When i run the test in the remote machine, i tried to access the application manually - one time it is very slow / after time it is very fast for some time.

Unfortunately we do not have access to the server to check the server's CPU memory utilization.

I am not expecting a very smooth curve in my results. But fluctuation should be in the acceptable range. Throughput varies from 2 to 60 to 2 What could be the reason for this huge fluctuation?

Additional Info:

  • Users : 300
  • Ram up period : 3000 seconds
  • The test was run for 2 hr 50 mins.
  • In 2 hrs 50 mins, It had sent 220000 requests in total.
1
I am not so sure from given information, it can be a possibility that because of 5 seconds delay in given sequence of requests, at one point of time all your users are doing some action(any request) and for next ~5 seconds system is idle or having very low load. That is why I think you are seeing this fluctuation. This again can vary based on rampup time provided but when you checked this manually also you might high response time when system is under load and might get fast response when threads are in waiting state.Nachiket Kate
To resolve this, I think you should minimize the delay and spread the users across realistic rampup time. Correct me if I missed something.Nachiket Kate
@NachiketKate, I had added some more information in the question. your theory does not look correct to me.KitKarson

1 Answers

0
votes

You didn't mention how many threads and you didn't mention what kind of timer.

It seems like the transactions are happening synchronously.

Check your ramp up period.
If you have 60 threads, ramp them up over 60 seconds to see what happens.

Or check your timer type, use a Gaussian Random timer or Uniform Random timer.