Web Testing for Reliability Improvement - Advances in Computers - page 206

F IG . 8. Top-level UMM for SMU/SEAS.
Lower-level models are also produced for the nodes “ /gradadmission/ ” and
“ /recruit/ ” in the top-level model. These models can be used to support our
hierarchical strategy for statistical usage-based testing.
The test sensitization and outcome prediction are relatively simple and straightfor-
ward, similar to that for testing based on FSMs in Section 4 . We can prepare all the
input and specify the anticipated output and transitions ahead of time for most test
situations. However, sometimes it would be hard to anticipate the input and the next
state. Under such situations, dynamic test cases may be generated in the following
way: From a current state, the transition or branching probabilities can be used to dy-
namically select the next state to visit, and sensitize it on the spot. Such dynamic test
cases also have their own drawbacks, primarily in the reduced system performance
due to the overhead to dynamically prepare and sensitize them.
The usage-based testing based on UMMs also yield data that can be used directly
to evaluated the reliability of the implemented system, to provide an objective as-
sessment of product reliability based on observation of testing failures and other
related information. Unique to the usage of UMMs is that the failures can be asso-