subject: Using Specifications In Performance Testing [print this page] At the time of design stage of the software development, performance and stability need to be determined.
If there are no precise definitions are set as targets to define a pass or failure, in this case software test may only recognize failure points, where the load has crashed the server or the performance is so poor that the customer-side connection has timed out.
It is critical that the specifications lay out exactly the behavior that is expected. Elsewise, Test cannot make objective claims and must dole out subjective judgments on the software's capacity for satisfying the client's requirements.
As s good software tester you should regard such scenarios:
If this application is an online retail Web site, you need stable uptime and cannot have users turned away because you didn't think they would wish to buy as much as they do.
If the application in question is a tool internal to the organization such as a time-reporting utility, then stability or performance might not be so critical. A providential quantity of users will be logging on and conducting a definite set of acts. In the case if something fails, it can be handled within the organization -people can just report their time later in the day.
If your organization is hosting Web sites, you may not expect to have high traffic because your customers might all be individuals and not organizations. What would happen when one user puts up one really useful article and a million users try to access it?
If this application is an Internet service provider, then the agreement may determine the quantity of hours per week users can be logged in or the amount of data that they can transfer. As software tester you would need to keep those determinations in memory when you perform software testing.
The specification needs to call out areas of potential concern based on the application's design. It is common known fact that the organization loses funds if it does not pay attention to such users' requirements. Without specifications that spell criteria out, data can be collected and be quite correct, but cannot be defined to be acceptable because there are no criteria by which to judge the data. If there are no definitions of the acceptable, desired, or unacceptable ranges, then Test cannot recognize a pass or failure rate, but only recognize points when the server stops responding or is unusable.
