Computing System Reliability Improvement by Using Unloaded Reserve

The goal was set to analyze computing system reliability which includes a number of servers modeled by the group of m one-type devices. Consideration made that non-failure operating time and recovery time of each device is spread according to exponential rule (with ζ and ν parameters accordingly). And time periods: operational and recovery, are mutually independent. Which allows to utilize model of "death and reproduction process". The system examined is serving requests coming from external Poisson flow with λ intensity, independent of system state. Requests are queued and are served by currently operational device during exponential rule distributed time with parameter μ. Cases analyzed: 1) all requests in the queue are stored and instantly transmitted to the device which is in reserve (interrelated solution); 2) all requests are dropped in case of operational device failure (independent solution). Formula obtained which describes correlation of time improvement of non-failure operation with interrelated solution comparing to independent solution (both with ν/ζ parameter). Correlation graphs are illustrated for different m-s, which can be used for preliminary estimation of computing systems reliability and number of reserve servers needed justification