Simulation of software testing as branched-cyclic technological process

In the paper the process of software testing is described as branched-cyclical technological process. The principal characteristic of the software testing process is cyclical process, namely that the program for the period of execution of each test cycle detected and corrected errors that were not discovered earlier. The authors proposed to build a model of the test to predict the cost of the testing process given its cyclical recurrence. We constructed a simple circuit of software testing process as a graph. We applied the algorithm of conversion the cycle graph of testing process to acyclic. We constructed Markov model for software testing process, including conversion cycle graph at the equivalent acyclic. The basis is taken heterogeneous Markov model. Authors applied the designed model on real data. We described Scrum technology of project management, that used for software development. The prediction is done through number of cycles required to achieve a given probability of absence of defects in software. To determine the probability of the states we used the results known passage of each module based on 50 iterations. We used preferred and undesirable conditions which can be operation after one cycle. The model made it possible to predict the performance of the process of software testing for three different modules. The proposed approach allows the prediction of the testing process for each software module that allows you to make informed decisions on each control operations in whole subprocess.