Influence Evaluation of the Routing Table Generation Algorithms for the Angara Interconnect on the Supercomputer Problem Performance

The qualifying work presents a performance evaluation of supercomputers based on the high-performance Angara interconnect using different routing table generation algorithms. In the work the author has integrated the collection of statistics on the volumes of data transfer between computational nodes into the routing table generation algorithm. Also, using the MPI library, synthetic tests based on communication patterns have been developed. Using the developed tests based on communication patterns and NPB tests, the performance of the Desmos supercomputer installed at the Joint Institute for High Temperatures of the Russian Academy of Sciences was analyzed depending on three routing table generation algorithms: based on search in breadth, based on search in breadth taking into account statistics, and a genetic algorithm. Keywords: interconnect, Angara, routing table generation algorithms, performance evaluation.