Generation of tests to determine the reliability of complex software.

Manual preparation of tests is advisable when a few values ​​have deep logical connections that are difficult to formalize. At the same time, manual preparation of tests is easier than developing a test generator. But such cases are relatively rare, and in most cases you need to automate the generation of tests.
Generators of software tests should provide the ability to create gestov in all the variety of dynamics of behavior and characteristics of objects of the external environment of this PS. Usually, only a small part of the tests is deterministic, and the main part reflects the stochastic and dynamic behavior of external objects.
When creating test generators, two fundamentally different approaches are used - integral and differential. With an integral approach, the basis is a formal description of the input and output information of the simulated object, as well as the functional relationship between the data at its input and output. The structure of the object and the processes that are realized in the operation of its components are not modeled and do not matter.
In the differential approach, detailed information is required on all the processes of functioning of the components of the environment. Differential or simulation models of test generators are based on this information.
In contrast to the natural experiment, the simulation of the external environment and tests on a computer has great potential for monitoring both the source data and all intermediate and output results of the functioning of the object under study. Software simulation of the environment allows
- to carry out long-term continuous generation of simulated data to determine the reliability of the PS operation in a wide range of changing conditions and parameters, which is often impossible for real objects;
- to expand the ranges of characteristics of simulated objects beyond the real-life ones;
- create test data corresponding to critical or dangerous situations of the functioning of environmental objects, which are impossible or dangerous to implement in-situ experiments;
- to ensure high repeatability of the simulated data under the given conditions of their generation and the possibility of stopping or suspending the simulation at any phases of the simulation environment.
Some even unlikely situations of source data values ​​are critical and especially important for the functioning of the entire system for which the MS is being developed. Selecting and simulating such situations allows you to work out PS in critical unlikely situations that are impossible or dangerous to create on real objects.
However, not always with software simulation, all the characteristics of the external environment are correctly and completely reflected, and separate control checks and software may be necessary in interaction with real objects.