On one decomposition algorithm for closed-loop periodic hypererlang service networks

A computing algorithm is developed for determining stationary probabilities of microstates in closed-loop $d$-periodic hypererlang service systems of complexity $0(n^3/d^2+(d-1)n^2/d^2+n)$ where $n$ is the number of network microstates. In comparison with a traditional approach of complexity $0(n^3+n)$ yields an asymptotic saving in time of about $d^2$ and requires less core memory.