Abstract:
The necessity of using hybrid solutions in creation of high-performance computing systems is substantiated. A brief description of the hybrid high-performance computing complex (HHPCC) of FRC CSC RAS is given and its enlarged block diagram is presented. The paper offers the methodical approach to estimation of reliability of HHPCC on the basis of which calculations of reliability of the allocated functional subsystems are carried out. Separately, reliability of the “computing infrastructure” of HHPCC (without peripheral elements) was evaluated. Recommendations for improving reliability of functional subsystems are given.