Scalability Testing

So far, we've discussed testing with a subset of the production web site. We begin by testing and tuning this subset of equipment, and continue until we reach our performance goals for this portion of the web site. However, we cannot assume the performance we achieve for the subset applies to the web site as whole. To achieve this confidence, we need a scalability test. During this test, we add capacity to our test environment and determine whether the web application actually takes advantage of this capacity. That is, does the web application scale to use additional resources?

Scalability testing usually doubles capacity at each step of the test. We start with a subset of our web site that supports some amount of user load. If we double the subset's capacity, we also expect to double the traffic the subset supports while maintaining the same response time. We call this linear scaling: As you multiply capacity, the throughput and users handled by the web site multiply as well, while maintaining the same responsiveness.

Scalability determines the behavior of the web site as a system, including its growth characteristics and limiting factors. These tests tell us how adding capacity at one point in the web site impacts the capacity of other components . We need this knowledge as we build our web site, as well as when we grow it in the production environment.