Testing is Rocket Science Not Brain Surgery

The Application is in Development

The performance-testing framework can be utilized early on during the Development Phase, by leveraging specific test types to address any particular areas of concern discovered during the assessment stage of the performance-testing framework. At this phase testing speaks to the application’s scalability and capacity by executing a profiling exercise. This proves to be an area of critical importance when discussing the recommended overall performance-testing strategy implementation. An introspective look at the type of testing that should be performed during the Development Phase yields the following observations, the first being that the developer unit-testing should be structured, time allocated organized, the second being that performance-testing should be performed in conjunction with this testing. It is the combination of these two forms of testing that will yield the highest degree of insulation from deployment risk at the onset. Typically, failing to utilize the performance-testing framework early on has resulted in performance-testing taking place at the end of the Devlopment Phase or later. This idea of only performance-testing an application that has passed the functional testing requirements usually yields little time to complete the performance-testing required, and making corrections should the post-test analysis identify a bottleneck or worse yet a poorly performing architecture overall, there will be little time to fix it. The test types most applicable at this stage are baseline, single-node, and component-level testing. This testing will provide the subjective and objective inputs need for the root-cause analysis effort later on.

The Application is in Test

Once the application has entered formal functional testing it should be complete enough to run the baseline, load, stress or volume tests against it. The models that represent the estimated usage of the system should be complete to allow for the creation of a battery of comprehensive real-worl scenarios. The tests should provide information to level-set expectations for the end-user response time and provide data for the top-down analysis effort. This can be followed up with component-level testing and the other test types native to profiling to investigate findings. This activity should be scheduled for multiple iterations.

The Application is in Maintenance

At this point in the application’s lifecycle any effort made to performance test should be a tuning exercise that would depend quite heavily on the tests used when adopting the profiling methodology. We should have a general idea of what areas are performing poorly and can begin component-level testing to investigate.