Abstract
This paper models a parallel system with multiple units performing and sharing a specified amount of work required for the system's primary mission (PM). The system operating environment is exposed to random shocks that may deteriorate or fail the units. To balance the risk of losses/damages caused by the mission failure and the amount of accomplished work, a dynamic task distribution policy (TDP) is proposed that, after the occurrence of each shock, determines the distribution of available units between performing the PM work and performing a damage reduction procedure (DRP). While this work belongs to the line of recent research on mission aborting, it makes advancement in the state of the art by considering the partial abort through the PM and DRP work redistribution among the available units (different from the PM full abort of existent models) and effects of external shocks for work sharing systems. The specific contributions made in this work include a discrete numerical algorithm for evaluating the PM success probability, expected fraction of the completed PM work, and expected total cost of losses (ECL) for the considered parallel system subject to the dynamic TDP. The TDP optimization problem is also formulated and solved, leading to the minimized ECL. Application of the proposed model and effects of several model parameters are demonstrated through a liquid supply system in the chemical reactor and through detailed analyses and optimizations of a multi-processor data processing system.