Editing Load Patterns to Model Virtual User Activities

The load pattern properties specify how the simulated user load is adjusted during a load test. Visual Studio Ultimate provides three built-in load patterns: constant, step, and goal-based. You choose the load pattern and adjust the properties to appropriate levels for your load test goals.

The load pattern is a component of a scenario. The scenarios, together with their defined load patterns, comprise a load test.

Load Patterns

Constant

The constant load pattern is used to specify a user load that does not change during the load test. For example, when you run a smoke test on a Web application, you might want to set a light, constant load of 10 users.

Constant Load Pattern Considerations

A constant load pattern is used to run the same user load during the run of a load test. Be careful about using a constant load pattern that has a high user count; doing so may place an unreasonable and unrealistic demand on your server or servers at the beginning of the load test. For example, if your load test contains a Web test that starts with a request to a home page, and you set up the load test with a constant load of 1,000 users, the load test will submit the first 1,000 requests to the home page as fast as possible. This may not be a realistic simulation of real-world access to your Web site. To mitigate this, consider using a step load pattern that increases gradually to 1,000 users, or specify a warm-up period in the Load Test Run Settings. If a warm-up period is specified, the load test will automatically increase the load gradually during the warm-up period. For more information, see Configuring Scenario Start Delays.

Step

The step load pattern is used to specify a user load that increases with time up to a defined maximum user load. For stepping loads, you specify the Initial User Count, Maximum User Count, Step Duration (seconds), and Step User Count.

For example a Step load with an Initial User count of one, Maximum User Count of 100, Step Duration (seconds) of 10, and a Step User Count of 1 creates a user load pattern that starts at 1, increases by 1 every 10 seconds until it reaches 100 Users.

You can use the Step goal to increase the load until the server reaches a point that where performance diminishes significantly. As load increases, the server will eventually run out of resources. The step load is a good way to determine the number of users at which this occurs. With the stepping load, you also have to monitor agent resources closely to make sure that the agents can generate the desired load.

Ordinarily, you should conduct several runs that have different step durations and step user counts so that you can obtain good measurements for a given load. Frequently, loads show an initial spike for each step as users are added. Holding the load at that rate allows you to measure system performance after the system recovers from the initial spike.

Step Load Pattern Considerations

A step load pattern can be used to increase the load on the server or servers as the load test runs so that you can see how performance varies as the user load increases. For example, to see how your server or servers perform as the user load increases to 2,000 users, you might run a 10-hour load test by using a step load pattern that has the following properties:

• Initial User Count: 100

• Maximum User Count: 2,000

• Step Duration (seconds): 1,800

• Step Ramp Time (seconds): 20

• Step User Count: 100

These settings run the load test for 30 minutes (1,800 seconds) at user loads of 100, 200, 300, and up to 2,000 users. The Step Ramp Time property is worth special mention, because it is the only one of these properties that is not available for selection in the New Load Test Wizard. This property allows the increase from one step to the next (for example from 100 to 200 users) to occur gradually rather than immediately. In the example, the user load would be increased from 100 to 200 users over a 20 second period (an increase of five users every second). For more information, see How to: Specify the Step Ramp Time Property for a Step Load Pattern.

Goal-Based

A goal-based load pattern resembles the step pattern but adjusts the user load based on performance counter thresholds versus periodic user load adjustments. Goal based loads are useful for a variety of different purposes:

• Maximizing output from the agents: measure the key limiting metric on the agent to maximize the output of the agents. Typically, it is CPU; However, it could also be memory.

• Reaching some target resource level, typically CPU, on the target server, then measuring throughput at that level. This enables you to do run-to-run comparisons of throughput given a consistent level of resource usage on the server.

• Reaching a target throughput level on the server.

In the following table, an example shows a goal-based pattern with the following property settings:

Those settings cause the Load Test Analyzer to adjust the user load between 1 and 100 during a test run in such a way that the Counter for % Processor Time of the WebServer01 hovers between 70% and 90%.

The size of the each user load adjustment is determined by Maximum User Count Increment and Maximum User Count Decrement settings. The user count limits are set by the Maximum User Count and Minimum User Count properties.

Goal-Based Load Pattern Considerations

A goal-based load pattern is useful when you want to determine the number of users that your system can support before it reaches some level of resource utilization. This option works best when you have already identified the limiting resource (that is, the bottleneck) in your system.

For example, suppose you know that the limiting resource in your system is the CPU on your database server, and you want to see how many users can be supported when the CPU on the database server is approximately 75 percent busy. You could use a goal-based load pattern that has the goal of keeping the value of the performance counter "%Processor Time" between 70 percent and 80 percent.

One thing to watch out for is if some other resource is limiting the throughput of the system. Such resources can cause the goal that is specified by the goal-based load pattern to never be reached. Also, the user load will continue to rise until the value that is specified for the Maximum User Count is reached. This is usually not the desired load, so be careful about the choice of the performance counter in the goal-based load pattern.

Virtual User Limits and Licensing

Visual Studio Ultimate lets you use up to 250 virtual users on a local load test run. If your load test requires more virtual users, or you want to use remote machines, you must purchase Visual Studio Load Test Virtual User Pack 2010. You can purchase Visual Studio Load Test Virtual User Pack 2010 where you purchased Visual Studio Ultimate. For more information, see Managing Your Virtual User Licenses for Load Testing with a Test Controller and Configuring Test Controllers and Test Agents for Load Testing.

Tasks

See Also

Other Resources

Creating and Editing Load Tests