Load tests are performed to measure how much load the system can take before it breaks. These types of tests are sometimes also called stress tests. Unlike in a performance test, there are many requests executed in parallel. What is measured is the average performance of all requests, while slowly increasing the number of requests to the system. In most cases, this will identify a breaking point, a specific number of requests per second at which the performance will suddenly decrease. This is the number of requests per second that the system can maximally serve. When executing a load test, gathering the average performance while increasing the maximum number of requests will often result in a graph like the following:

This graph shows why it is important to know the breaking point of an application: too much load might crumble a system unexpectedly because of the sudden nature of the change in response times. Knowing where this point is allows operators to act before this point is reached in a production environment.

At the end of this chapter is a link to an online Microsoft lab for developers to practice load testing.