A novel technique of reliability estimation for a stochastic-flow computer system from QoS perspective

From the quality of service (QoS) perspective, this paper proposes a stochastic-flow network model to estimate the capability of a computer system with considering both edges and nodes may fail. In such a system, edges (transmission media) and nodes (transmission facilities) have various capacities due to failure, partial failure, or maintenance. Thus, a computer system consisting of such edges and nodes also has several possible states and it is a so-called stochastic-flow network, namely stochastic-flow computer system (SFCS) herein. To guarantee that the SFCS keeps a good QoS, the maintenance action is taken when the SFCS falls to a specific state such that it cannot provide sufficient capacity to meet demand. An algorithm is developed to estimate the system reliability as a key performance index (KPI). Based on the KPI, the system supervisor can conduct the sensitive analysis to improve/investigate the most important part in a large SFCS.