Scalability testing | HA testing

Soak, endurance, or reliability testing

Reliability testing is done around expected peak capacity (P) and higher. Usually, the testing should be conducted close to required minimum business capacity (B) but not more than the maximum system capacity (C).

The key factors of reliability testing is the duration for which the test is run as well as the variable workload to mimic typical day-to-day activities on your site.

Reliability runs are usually run for days and can be half a day, one day, and so on, generally going up to seven days. A three-day reliability run is very common, since this is usually long enough to start showing symptoms of many performance problems, should they exist. A three-day test also gives you an opportunity to start it on a Friday, observe the test for a little while, then leave it running for the weekend, and then come back on Monday and observe the test in execution for another little while before it completes.

Figure 20-4 gives an example of a three-day reliability test. The workload curve is not a straight line, but is modulated as shown in Figure 20-4. In addition to varying the workload from the shoppers, extra workload of the likes generated by DB clean, WCA, massload, Sales Center activity, MQ, Staging, and Authoring server will be added as well. This utilities can be executed manually or, preferably, automatically.

xxxx