1994: Techn. Report No. I-09. BTU Cottbus, 1994. Available at http://www.informatik.tu-cottbus.de/~wwwdssz/publications/papers.html.
Abstract: Starting from a classification of software validation techniques an innovative Petri net based methodology is proposed to integrate qualitative and quantitative analysis methods of parallel software systems. The approach combines qualitative analysis, monitoring and testing as well as quantitative analysis on the basis of a net-based intermediate representation of the parallel software system under consideration.
The validation of qualitative properties comprises two steps. At first, the context checking of general semantic properties is done by a suitable combination of static and dynamic analysis techniques of Petri net theory. Afterwards, the verification of well-defined special semantic properties given by a separate specification of the required functionality is performed.
The validation of quantitative properties is based on so-called locally Markovian Object Nets (MONs) which are obtained from the qualitative models by property-preserving structural compression and quantitative expansion. Here, the frequency and delay attributes necessary to generate quantitative models (i.e. the MONs) are provided by the monitoring and testing component.
Besides the provision of the quantitative attributes according to the user-driven time abstraction level, the net-based testing method supports a systematic test of parallel systems. Techniques to derive automatically dedicated test suites and to measure the test coverage obtained are important features of this systematic testing. The basic ideas are highlighted by a small running example.
Keywords: Parallel software engineering; process-oriented imperative languages; software validation; static analysis; testing and monitoring; performance evaluation; dependability; formal methods; Petri nets.