Analysis of problems in the research area of software reliability: a lot of stages and architectural aspect

In the survey, it is shown that estimation and reliability analysis should be performed at each step, create a failover. Due to the fact that there is no single approach, methods and parameters for estimating the reliability of Software, there are many models and algorithms, tested virtually, but has several disadvantages. The main drawback of all models is "specialization ". Models are always tied to any one phase of software development, and sometimes have a purely theoretical value due to the fact that they must be used the data obtained at other stages of development, and the authors of such models usually consider this fact of secondary importance. The most promising approach is aimed at creating some hybrid model, which includes all the advantages of a well-validated models and algorithms. Existing models and algorithms need to be modified so that the output of one model became the input for the other; to provide a unifying parameters, units of measure conversion, normalization of data and satisfying the current requirements and trends, such as object-oriented analysis and programming. From this review we can see that it is very important to use as much a priori information before beginning each phase of creation, as this will provide the most appropriate and accurate results. The analysis of methods, models and methodologies in the field of creation of reliable, fault-tolerant (Software) is carried out. Attempts to systematize and unify the existing methodology are made. The multiple steps involved in the creation and architectural aspect within the entire life-cycle fault-tolerant software are taken into account.