Reliability of a Cluster of Duplicated Computer Systems with the Criticality of Functional Requests to Waiting

Abstract

High reliability and fault tolerance of computer systems at high performance and low delays of executing the required functions are achieved by consolidating the data processing and storage resources through the use of cluster systems. Ensuring the reliability of cluster architecture computer systems while accelerating their reconfiguration after failures is accompanied by replication and migration of virtual machines. A approach is proposed for the justification of building a computer system with a cluster architecture, composed of duplicated computer nodes. Studies aimed at increasing the readiness of the cluster to timely execution of real-time requests while ensuring a stationary service traffic mode have been carried out. The work of duplicate nodes in load balancing mode and duplicate calculations is considered. In the first mode, requests are evenly distributed among the functional nodes (servers). In the second mode, replicas of the request are created, each of which is executed on one node of the backup system. For control, the calculation results are compared at check points. If there is a mismatch, a repeat calculation is possible.