Resource allocation decision model for dependable and cost-effective grid applications based on Grid Bank

Distributed and parallel computing systems such as grid- or cloud-computing have been widely applied, studied and developed for various large-scale computing requirements in cross-administrative domains. However, two important issues, the economy of computation and the dependability of service-oriented computing, have not been thoroughly deliberated. In this regard, we formulate a grid resource allocation decision model that includes (1) a service reliability assessment, which derives the computing dependability from the universal generating function methodology (UGFM), and (2) a virtual payment assessment, which appraises the service expense for the contribution of each resource based on the expense function. For the near-optimal (or optimal) non-dominant solutions (service expense and service reliability), this paper presents two bi-objective soft-computing techniques, PC-GA (genetic algorithm) and PC-PSO (particle swarm optimization), where the Pareto-set cluster (PC) contains elite-selected and reborn mechanisms to generate new non-dominant solutions and strengthen the optimization effectiveness of the Pareto frontier. Finally, a virtual grid system is provided as a case study to illustrate the performance of two optimization methodologies and to analyze the pros and cons in terms of different resource allocation decisions.