Multilevel behavioral testing of distributed information systems

Abstract

Operational features of modern distributed information systems significantly increase the requirements for the reliability of their functioning and tighten the time limits for its restoration. High reliability of work is provided by various tools, in particular, offline and online testing, but often limiting the real time of the operation of these systems. Well-known testing methods use hardware accelerators, decomposition and parallelization, artificial intelligence technologies, but their capabilities do not remove the advisability of developing new models and methods, in particular, high-level ones, based on a special analysis of system behavior. The article presents a multi-level method of online behavioral testing of distributed information systems, based on hierarchical Petri nets, single-level model and method, and also hierarchical model of online behavioral testing. The multi-level method uses evolutionary optimization and has the features of parallel multi-level organization of compatible check evolutions of individual levels, which develop according to the principle of a “wave” algorithm in the background relative to the main functioning of the distributed information systems themselves. To determine the complex testing criteria in the check evolutions of the multilevel method, a hierarchical system of compatible functions of fitness and the choice of objects in the level evolutions of the hierarchical model is proposed. The functions offer completeness estimates for positions, transitions and arcs of Petri nets, linear and nonlinear graph behavior structures, the action of functions is extended to check primitives and fragments. The hierarchy of fitness and choice functions made it possible to introduce estimations of completeness of testing used to optimize evolutionary search in a multi-level coverage of necessary checks. Hierarchical decomposition, evolutionary approach and hierarchy of optimizing fitness and choice functions of a multi-level method made it possible to reduce time and increase the flexibility of organizing online behavioral testing, taking into account the features of distributed information systems.