A Technique of the Distributed Information Systems Control Method Choice under the High Network Dynamics Conditions

Abstract

Purpose of research. The purpose of this study is to select a method for managing a distributed system, which, based on known parameters, would reduce the consumption of resources of computing devices. The resource of an information system is understood as the probability of failure-free operation (reliability function), which degrades over time for each node, the faster, the higher its workload.Methods. Considering that with high dynamics of the edge layer of the network, the frequency of system reconfigurations becomes relatively high, and the need for reconfigurations is unpredictable, reducing the total time spent on reconfigurations makes it possible to increase the time spent on solving functional computational problems of the system and thereby reduce the load of nodes. The reconfiguration time can be reduced both by reducing the time for detecting a failure in a distributed system, and by reducing the new configuration forming time. In this paper, a method for reducing the time of the failure detection is considered. The analysis of the applicability of system control methods (centralized, with a distributed leader, decentralized) is based on the obtained analytical estimates of the time the system detects a failure under control conditions using one method or another. A numerical experiment makes it possible to identify areas of system parameters, where it is preferable to use the method with a distributed leader.Results. The main result of this work is a methodology for choosing a method for managing distributed information systems in conditions of high dynamics of the network infrastructure, focused on reducing the consumption of resources of computing devices.Conclusion. System reconfiguration time can be shortened by choosing the most appropriate control method. Thus, the time allotted for solving the functional tasks of the application increases, the workload of the computational nodes decreases, and, therefore, the FBG values increase over the planning horizon.