This three-part paper analyzes existing approaches and methods of organizing failure- and fault-tolerant computing in distributed multicomputer systems (DMCS), identifies and provides rationale for a list of issues to be solved. We present the concept of fault tolerance proposed by A. Avizienis, explicate its dissimilarity from the modern concept and the reason for its inapplicability with regard to modern distributed multicomputer systems. We justify the necessity to refine the definition of fault tolerance approved by the State Standards, as well as the necessity to specify three input parameters to be taken into account in the DMCS design methods: permitted fault models, permitted multiplicity of faults, permitted fault sequence capabilities. We formulate the questions that must be answered in order to design a truly reliable, fault-tolerant system and consider the application areas of the failure- and fault-tolerant control systems for complex network and distributed objects. System, functional, and test diagnostics serve as the basis for building unattended failure- and fault-tolerant systems. The concept of self-managed degradation (with the DMCS eventually proceeding to a safe shutdown at a critical level of degradation) is a means to increase the DMCS active life. We consider the issues related to the diagnosis of multiple faults and present the main differences in ensuring fault tolerance between systems with broadcast communication channels and systems with point-to-point communication channels.
The first part of the work mainly deals with the analysis of existing approaches and methods of organizing failure- and fault-tolerant computing in DMCS and the definition of the concept of fault-tolerance.