Model-based Technology of Automated Performance Testing

Abstract

Abstract The Technology of the scaled-down experiment for evaluation of the performance characteristics of wide class of systems for business-process automation has been developed. The Technology objectives are to provide adequacy of the results of particular experiment and its components: problem definition, source data and analysis of results of the experiment. The Technology is based on the set of metamodels that define main components of the experiments. Technology provides technique for adaptation of metamodels to the parameters of any particular automated performance testing experiment. 1. Introduction The major works in the area of Model-based testing (MBT) are mainly aimed at the use of various statistics programs to generate tests data or to estimate tests results at automated functional testing. Detailed analysis of these works as well as creation and usage methods at MBT is given in the papers of Whittaker J.A. [5], Petrenko А. [6] and others. As for functional testing, the use of models, reflecting tested programs statistics properties (structure, data flow) is highly effective. Model dimension can reach up to hundreds and even thousands of vertexes. In this work the model-based approach is treated in the connection with information system (IS) performance testing. Within its framework application software testing in IS is carried out in functioning environment, that is, in the environment, including DBMS, data base, system software, hardware. Few works are devoted to model-based performance testing methods. Model-based methods are most frequently applied in case a large number of virtual clients generation is involved in the course of web-systems and servers system testing in order to define their workload capacity [7]. Service state change structure, represented by web-system taking into account service realization probabilities and their sequence, is being modeled. There are also papers on model creation for networks analysis at their performance testing. [8]. Such a universal model, for example, allows to unite different testing tools into unified technology on the basis of several types, so called Universal Probe (UP), that define architecture, system element interaction policy and resources constraints. However, as it is seen from practical testing of critical IS, the above said models do not cover performance testing planning and evaluating adequacy of results. IS performance testing in the operational process is mainly aimed at