Forum Discussion

Dave_Meibusch's avatar
Dave_Meibusch
New Contributor
17 years ago

Load tests timings : rounded to milliseconds?

I'm load testing my services on pretty quick new hardware. Typically the web service request elapsed times on the server are in the 1-2 millisecond range.

However, soapUI load test runner (command line version, 2.5.1) appears to be rounding all the response times to the nearest millisecond (I assume).

It reports an average elapsed time per request of less than the time that the server records to process the request. soapUI appears to be recording a time of 1ms for each request (and occasionally 2ms), hence an average just over 1ms. The server accurately records the requests as averaging about 1.72ms. I've double checked my server code - it looks correct.

In the soapUI manual there is a section comparing JMeter to soapUI. It specifically notes that soapUI uses System.nanoTime() rather than System.currentMillis(), but it would appear that it may later round these values to milliseconds.

Can anyone confirm that this is the case?

4 Replies

  • Hi Dave,


    I can confirm that soapUI measures time in nanoseconds but sum times are truncated to milliseconds prior to displaying them. In other words, the times measured by soapUI in your case were 1.72ms<2ms, hence the unexpected result.


    Cheers!
    /Nenad Nikolic a.k.a. Shonzilla
  • Thanks for the confirmation.

    Has this been ticketed as an issue for improving in future releases?

    thanks,
    Dave
  • Hello Dave,

    we will definately look into improving this in the future, thank you for bringing it to our attention.

    Regards,
    Dain
    eviware support
  • Two workarounds for this issue possibly:

    1. Groovy scripts to record times
    2. Java extension (test listener impl)

    But neither are much good it appears as timings outside the test execution include too much "extra" time. I really need an aggregate of the execution time *excluding* send/receive encode/decode time.

    Ah well, maybe a future release.