Performance Testing for Dummies

What is Performance Testing?

Performance Testing is defined as “general testing to determine the effect of load on a particular application.”

This means testing an application with Virtual Users to test the Response Times (time taken for the application to respond to a given input) and Resource Utilisation.

Why Performance Test?

To identify points at which the end users experience becomes undesirable.
When the response times increase to levels which inhibit the use of the application, the end users experience will be generally unpleasant. This can lead to poor customer satisfaction, lost revenue and reputational damage.

To identify points where the system architecture responds in a manner which is undesirable.
This allows one to identify failure scenarios and possibly confirm the need for more Capacity, allowing the Capacity plan to be verified.

To identify issues with the software which are obscure.
These issues are small, hard to detect while functionally testing yet could have devastating effects on performance. These include thread leaks and memory leaks.

To identify performance optimisation changes are working.
By comparing new results against a baseline or previous test, one can ensure than changes made to optimise performance are having the desired effect.

What sort of Tests?

Here are 3 examples of tests used to test an application, each tests performance but focusing on only one can lead to undetected performance issues that only surface once the application is “live”. A broad view with testing spread over a whole range of scenarios is best, with particular focus applied to areas of concern.

Breakdown Test
Used to break the application into its constituent sections and identify the areas of the application with poor performance.

Ramp Test
Used to test one constituent part or path through the application with increasing virtual users over time; allowing response over load to be observed.

Soak Test
A test lasting an extended period of time, used to test an application’s response over continual load.

How to performance test?

This is entirely dependent upon the application and requirements. Generally though the application is hosted within a given test environment, which models the live environment in a scalable manner. Herein virtual users are injected onto the application and data captured from the application and the test environment.

At Capacitas we do not recommend any specific program or vendor, being completely independent we assess your situation and suggest the correct tool for the job.

When to performance test?

As soon as possible, to understand what the limitations of the environment and to gain an understanding into whether there may be environmental issues.

Once the first steps of the application are working the initial testing should be carried out, if the responses are very far from the requirements a change is easy to accomplish and relatively cheap.

The cost of resolving issues once large development is complete can be very expensive; it is worth bearing this in mind before jumping into full scale development.

Final Thoughts

While performance testing is generally second to getting it to work, customers are fickle and a poor performing application can lead to lost revenue. A small investment initially can reap rewards in the long run.