Hi Chan,
Here are the steps:
1. Launch a test with N virtual users.
2. Increase the number of virtual users and launch it again. Make sure the new test is launched in the same group (see the
Load Testing Analysis Page help topic for more information).
3. Repeat steps 1-2 until you get the needed number of test runs. Each test run will represent a single point in your dependency graph.
4. On the Graph tab of the Load Testing Analysis page, build a graph which will display User Count on the X axis and Avg Process Time on the Y axis (see the
Working With Graphs and Charts help topic for more information). The Y axis will not show the response time directly, but the characteristic (which is average virtual user life time (in seconds) during the test run) strongly depends on response time and this will allow you to make conclusions on dependency between the number of virtual users and response time.
You can use server performance counters to measure throughput. To learn how to do this, please see the
About Host Performance Counters help topic.
Utilization can also be measured via the corresponding server performance counters (Memory, CPU, Disk I/O).
2. How to determine min. bandwidth required based on the testing result?
|
You can use the Performance column for that (see the
Load Testing Log Page help topic for more information). For example, you can reproduce the required conditions (e.g. N users take M seconds to complete the desired action) and use the max/min/average performance value for that conditions to make conclusions on the required bandwidth.
3. At which point (how many users) performance started to decrease?
|
If you want to examine how your application reacts when the number of virtual users increases, I recommend that you use the ramp up feature. Please see the
Simulating Load Increase help topic for more information. As an alternative, you can perform multiple separate test runs with different number of virtual users and measure the needed characteristics (depending on your "performance decrease start" conditions).
What is the max. # of user can system handle?
|
The specific steps depend on your scenario. As in the previous case, you can repeat the same test with different number of virtual users and measure the needed characteristic(s) to check how your system reacts to the increasing number of virtual users. The characteristic you need depend solely on your requirements: you can increase the number of virtual users until at least one error occurs and/or until the time necessary to complete the test exceeds some value, etc.
4. What is the time taken for each user in the test? I am not looking for performance time. Instead, I want to know avg process time for each user.
|
A virtual user's life time is displayed in the Time (s) column of the Users panel in your load testing log.
5. Which request always take longest time to process?
|
Here's what you need to do:
1. In your load testing log, select a virtual user in the Users panel.
2. Select a connection in the Connections panel.
3. Right-click somewhere within the Requests panel and select the "Show Summary" context menu item.
4. Right-click the summary panel below the "Response Time (s)" column and select the Max context menu item. This will allow you to see the max response time for the current connection established by the current user.
5. Enumerate all connections of all virtual users to get the max response time value.
