What is the volume of evidence?
At first glance it might seem that the evidence should be high volume by model V, if they take the above definition. It is certainly under non-functional requirements testing & performance testing. volume tests can (& should) be used in component testing. This would be a close correlation with the code, so at this level could be connected to Dynamic Analysis. Volume tests were also carried out (generally) as part of acceptance testing by users. Stress testing is closely related, as it tries to find out how program will behave beyond their specified limits.
Why conduct a volume test?
1. for errors
2. Give reliable information on the status of the component, in which business decisions can be taken.
As for the flaws to be found through volume tests are those that, if program behavior deviates from expected for a given volume of information. Thus, a banking technique will be tested for faults in much larger volumes of information, that the little retail program. A bug that only manifests itself on a table with a million records is not relevant to the retail program, but will be picked up by the testers of the bank. Reliable information about how program will behave is essential. During the dotcom boom plenty of web-sites were launched without knowing what the effect would be that the database backend grew exponentially. Of work plenty of accidents as a result.
Why component-level test?
This is because then they can see how the code behaves, & confirm that the component won't be a bottleneck & slow down the technique as a whole. (You can also use plenty of technique resources.)
For example, a window often used window object is populated with information by calling an object database that runs a complex SQL query. Assuming that the component is tested against the boards with only 4-5 logs. Of work, back in an instant. Everything seems fine. Is then integrated with the window, & the test technique. Again, everything seems ok. It is only when the application is user acceptance (or even gone live) compared with the 100,000 records, it was discovered that SQL was not well optimized & tables are not indexed. Therefore, it should have been tested at the component level. Some methodologies such as RUP early trials suggest. Therefore in the early stages (beginning & elaboration), volume of tests could be conducted to confirm that the core architecture is to proceed with three.
Who will be tested for volume?
Developers across customers & finish users can do. The tests can be outsourced to a testing laboratory that specializes in performance testing.
Using the volume of evidence? Since the test volume can be introduced in to anywhere, from tests of the components of the user acceptance test, the range of persons who can perform is equally great. Volume testing needs three things.
- The clear outcomes of how the program behaves in a positive level of information.
- The information, & plenty of it. The expected behavior at different levels should be in the documentation of the specification.
Ideally, this will say something like "the customer details will be returned back on the screen within 3 seconds, from a database with 1 million customer records." This gives the tester a reference point based on a test case. As with all tests, the relevant documents should be kept showing the inputs, outputs information on the other, for the purification & the potential for audit purposes. The second requirement for the information, the needs of both real-life information and simulated information. Usually, the real-life information will come in the form of a database of customers who have had private information such as names & account numbers scrambled. Records also can be created from scratch using automated tools or by adding rules directly to the SQL database.