locked
Where do I install the Visual Studio Controller and Agents? I want to run the load test in Visual Studio 2010 ultimate. RRS feed

  • Question

  • Hi Team,

    I am new to performance testing, using Visual Studio 2010 ultimate . Can we install and configure the Test Agents and Controller in the same machine where the Visual Studio 2010 ultimate has been installed. Will it cause any impact on the load test with 250 virtual users if i get it done on the same machine ?please advice me elaborately with the best possible solution.

    Thank You.

    Wednesday, February 5, 2014 7:05 AM

Answers

  • Visual Studio Ultimate can create and run load tests without needing agents and controllers; Visual Studio includes all the required functionality.

    Each virtual user in a test run consumes some CPU time and needs some memory in the computers running the tests. The reason for agents and controllers is to avoid overloading the CPU or runnning out of memory when lots of virtual users are active. Configuring agents and controllers in the same computer as Visual Studio will only decrease the resources available for virtual users. (I am not sure you can install and run agents and controllers on the same computer as Visual Studio.)

    Agents and controllers are normally only needed when large numbers of virtual users are required. The maximum number of users that Visual Studio can handle from one computer without adding agents depends on several factors including: power and memory of that computer, test case duration, test case complexity, think times, numbers of plugins and extraction rules executed, etc. We use 250 as a first estimate of how many virtual users one computer can handle.

    For your test with 250 virtual users I would suggest just use Visual Studio without adding any agents or controllers. Monitor the performance of the computer, using the load test graphs, as the tests run. If the computer seems overloaded then add an agent computer.

    Regards

    Adrian

    • Proposed as answer by Shiv_p Wednesday, February 5, 2014 2:37 PM
    • Marked as answer by Sanu CS Thursday, February 6, 2014 9:43 AM
    Wednesday, February 5, 2014 9:35 AM

All replies

  • Visual Studio Ultimate can create and run load tests without needing agents and controllers; Visual Studio includes all the required functionality.

    Each virtual user in a test run consumes some CPU time and needs some memory in the computers running the tests. The reason for agents and controllers is to avoid overloading the CPU or runnning out of memory when lots of virtual users are active. Configuring agents and controllers in the same computer as Visual Studio will only decrease the resources available for virtual users. (I am not sure you can install and run agents and controllers on the same computer as Visual Studio.)

    Agents and controllers are normally only needed when large numbers of virtual users are required. The maximum number of users that Visual Studio can handle from one computer without adding agents depends on several factors including: power and memory of that computer, test case duration, test case complexity, think times, numbers of plugins and extraction rules executed, etc. We use 250 as a first estimate of how many virtual users one computer can handle.

    For your test with 250 virtual users I would suggest just use Visual Studio without adding any agents or controllers. Monitor the performance of the computer, using the load test graphs, as the tests run. If the computer seems overloaded then add an agent computer.

    Regards

    Adrian

    • Proposed as answer by Shiv_p Wednesday, February 5, 2014 2:37 PM
    • Marked as answer by Sanu CS Thursday, February 6, 2014 9:43 AM
    Wednesday, February 5, 2014 9:35 AM
  • You can locally run 250 virtual users in 2010 Ultimate. For more than 250 virtual users, you will need Visual Studio Load Test Virtual User Pack 2010 licenses.

    As Adrian suggested, sizing the test agent depends on several factors.

    But generally speaking, assuming your application is not so rich (web 2.0), you can manage to load for 250 locally without the need for test agent.

    I suggest you first create a web test and then run it for 1-user to observe the memory usages.

    You can then use this as bench mark and calculate the memory required for 250 users.

    I used 3-test agents each 4GB RAM to load test CRM application for 1000-users.....so 4GB machine should suffice your requirement.

    Hope this helps!


    Wednesday, February 5, 2014 2:35 PM
  • Thank you for the response and suggestions , I shall try it!

    My Visual Studio 2010 ultimate System has a Processor : Intel(R)Xeon(R) and RAM : 8 GB.

    Manually when i test my application, the time out is set for 20 minutes. When i configure my load test for 250 users  for 20 minutes, I set the Thinktime as 30 seconds and Timeout as 1200 seconds respectively for each of the pages recorded in the Web test in the Load test Editor page. Warm up duration=1 minute, Sample rate=10 seconds. Is it right what I am doing?

    My another concern is:

    While adding a load test, in the New Load test wizard, "use normal distribution centered on recorded think times" what does this actually mean? Think time between test iterations = ? Which 'Test Mix' model is advised for the load test patterns?

    Thanks in advance :)!!!

    Thursday, February 6, 2014 10:11 AM
  • A web performance test typically has a recording of a users interaction with a website. It will contain several requests that are sent to the web site. The "think times" correspond to the pauses between (a) receiving and displaying data from the web site and (b) the sending of the next request that a person would make when using the web site. Without think times the test would run unrealistically fast and thus impose an unrealistic load on the web site. The "use normal distribution ... think times" refers to altering the think times randomly to better reflect the speed of different users. The "normal distribution" parts means that most of the random changes will be small, but a few will be larger.

    The think time between test iterations provides an additional pause between the end of one test execution and the start of the next.

    Regards

    Adrian

    Thursday, February 6, 2014 6:06 PM