Determination of exact upper estimates of time taken to perform complex sets of problems in control, parallel computing systems

The actual problem for prediction of the reliable completion of the user-prescribed complex of interconnected jobs (CIJs) – the collection of interdependent program modules with random times of their realization – is solved in control, parallel computing systems by determining the exact upper estimate of the distribution functions of the time taken to complete the CIJs on condition that the number of computing devices of the system is no less than the coefficient of parallelism of the CIJs. Usually, a rather cumbersome obtaining of similar estimates is replaced by the construction of an aggregate of multiple integrals evaluated by the Monte Carlo method.

Presented by the member of Editorial Board: P. P. Parkhomenko