Riznyk O., Vynnychuk R.
Abstract. Clusters and distributed systems allow for error tolerance and high performance through shared use of the load. When all computers work, we would like to distribute the load equally among computers. When one or more computers are broken, the load on these computers should be redistributed to other computers in the cluster. Redistribution determines the recovery scheme. The recovery plan should keep the load as optimal as when even the most unfavorable combinations of computers are turned off, that is, we want to optimize the behavior of the worst case. In this work, we find new regeneration schemes based on so-called IRB and ideal ring loops and synthesize the corresponding combinatorial structures. They are optimal for many cases of recovery schemes.
Keywords: сluster, distributed system, Golomb ruler, ideal ring bundle, recovery scheme