The reliability model of the restored cluster is proposed, which provides for physical and informational recovery of memory nodes, carried out on the basis of replacing the failed memory node with a node in which the results of the last backup are previously entered, with further updating of information when replicating data from working memory blocks containing up-to-date information. If all memory nodes with up-to-date information have failed, then using the system to perform the required tasks until the information is fully restored is fraught with risks of erroneous request servicing. Options for reducing these risks are considered.

Vladimir Bogatyrev

Anatoly Bogatyrev

Readiness for timely execution of requests in fault-tolerant clusters with information recovery based on replication and backup