The Research Of Reliability Test Approach Based On Markov Usage Model For Real-time Software

With the rapid development of computer technology,software applications have infiltrated various fields of social life,and the impact of software failures has also been increasing.As the most important user-oriented quality feature,the software reliability level has already become one of the most important authentication indicators for users when the software is released.Software reliability test is an important way to ensure and improve the software reliability level.There are a great variety of models for reliability testing,Markov usage model is a very representative and widely used one.Reliability testing based on Markov model generally starts with describing the use of software by models,then followed by the test execution,test results analysis and other steps.It can be said that the appropriate and effective model is the key to a successful test.Through in-depth study of relevant literature in reliability testing,the existing research basically solved the process of converting from UML model to Markov usage model,that is,the corresponding Markov usage model can be analyzed from the source code of tested software.For the time constraints existing in the real-time software system,there is also some conversion and representation.However,few studies have mentioned how to quantitatively calculate the impact of time constraints on test case generation.Traditional reliability testing based on the Markov usage model usually generates test cases only by usage.However,time constraint is also an important attribute affecting the software reliability especially for the real-time system.Based on the Markov time usage model,a feasible framework for time-constrained optimization is proposed in this paper,on which an automatic model optimization algorithm is proposed to maximize the model state entropy.The optimized model can reflect,to a certain extent,the time attribute of the tested software system.In addition,although some corresponding handling strategies have been proposed for the redundant states and equivalent states in model,the cyclic sequences haven't been given much attention.In fact,there often some loop fragments in Markov model while created according to the usage rules or automatically generated from the UML model,which may cause problems such as excessively generating test cases,wasting test resources,or even getting run-down if no treatment or limitation.Aiming at these defects in model,an optimized test case generation algorithm is proposed.The first step is to shrink the cyclic segments in original model,and a special tree data structure SCC-Tree is used to store and mark its inner structure.As for the acyclic model,a complete set of test cases is generated by a certain combination strategy.Experimental results show that generating test cases with the algorithm proposed is more efficient and stable than traditional one on the basis of meeting the path coverage criteria.Finally,the reliability testing prototype system based on Markov model is designed and implemented with the proposed algorithm.The example of the UAV sub-module is used to verify that the system is simple to use and the method is effective.