Results 1 to 5 of 5
  1. #1

    Reg reliability testing math......or math that ext

    I am a performance tester and i need to know more about extrapolation of performance test result. Do you have any math to extrapolates the results of performance test.

    I also need to suggest a better plan for reliability and extrapolation of performance test result.

    I have done the performance testing for 500 Vusers, but now our client is asking to certify there product on basis reliability factor. Please help me to find out how can i certify or say to client that there application or software is this much reliable?

    Please let me know is there any tool available for checking the reliability of application or software on the basis of performance test data.

    have a one more requirement from of the client to Extrapolation of Load Test Results.

    Here the idea is for ex: if we have license for say 500 Vusers but we need to certify the application for 1000 Vusers using Extrapolation of Load Test Results.

    Can you please help me how to do Extrapolation of Load Test Results? and which all factors needs to be consider in doing that.

    Please do the needful.
    In Seek of your help.

  2. #2

    Re: Reg reliability testing math......or math that ext

    This is not a trivial exercise and I am not sure there is a good solution for it.

    TeamQuest and HyPerformix have products oriented for capacity planning. They are not necessarily easy to use. As LoadRunner, they require knowledge and experience in order to extract the best out of them. I think you can acquire a service from both companies in order to address your requirement.

    Alternatively, you can run a test in LoadRunner which is aimed to understand how the application performs under gradual growing workload. Meaning, if your test environment can sustain workload generated by 500 concurrent vusers, build a test from 5 steps (100 Vusers, 200, 300, 400 and 500). Each step should have steady state so you can deduce how the application behave under each and then once the test is over you can compare the performance pattern between these steps. This might teach you about the performance in larger workloads, but not in a high probability.

    Shlomi Nissim
    LoadRunner and Performance Center Functional Architect
    HP Software and Solutions

  3. #3

    Re: Reg reliability testing math......or math that ext

    Hi Shlomi,

    Thanks for your reply... but the thing is here i need to extrapolate the results and i knew that if i follow the gradual growing workload and noted down the parameters for all the runs but still i cant asure client that i can certify the application for 1000 Vusers. Please suggest whether its best practise to do that.

    And do you have any idea for reliability performance testing approach?

  4. #4

    Re: Reg reliability testing math......or math that ext

    There are too many parameters that may impact your result in case the application is tested on different environment than the place in which it is going to be deployed (HW configuration, tuning of different tiers, database volume, etc...), hence I am not sure it is practical to explain it in the forum.

    There are methods for extrapolation, none guarantee 100% accuracy/certification.

    If you measure average CPU utilization and response time, for each steady state, in the gradually growing workload approach, you can come up with a linear formula that pretends to predict how the application wil behave in bigger workloads. This is pretty simple mathematics, but since it is linear formula it can't provide high confidence

    sorry for not being more helpful
    Shlomi Nissim
    LoadRunner and Performance Center Functional Architect
    HP Software and Solutions

  5. #5

    Re: Reg reliability testing math......or math that ext

    Performance is not simply a matter or arithmetic progression, which is why the tools noted above exist. One does not simply use up network, disk bandwdith, CPU or RAM at the same percentile rates as your system comes under load. Different mixes of users can also cause different resource utilization levels. At some point one resource may be constrained and this will cause response times to grow even if all of the other resources are clean.

    Graphically when this happens the response times look a lot like a hockey stick as they turn from a nice arithmetic progression to a logarithmic one. Modelling tools will help you to run "what if" situations to see where you might hit a limit, but tven these manufacturers will tell you that at some point you need to validate your model by actually running the larger test.
    James Pulley

    Replace ineffective offshore contracts, LoadRunnerByTheHour. Starting @ $19.95/hr USD.

    Put us to the test, skilled expertise is less expensive than you might imagine.

    Twitter: @LoadRunnerBTH @PerfBytes



Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
BetaSoft Inc.
All times are GMT -8. The time now is 03:55 PM.

Copyright BetaSoft Inc.