A User Satisfaction Level Based Multi-objective Optimization Method for Grid Task Scheduling

Abstract

A user satisfaction level model, which focuses on the real feeling of grid users, is proposed in this paper. The model can integrate not only multiple-aspect demands of different grid users, but also their preferences in different aspects. An abstract user satisfaction level function and its properties are also defined in the model. Based on the model, an effective grid task scheduling algorithm Sat Opt is provided and a multivariate Gaussian function is chosen as the practical user satisfaction level function, The Sat Opt algorithm is a three-stage heuristic multi-objective optimization algorithm and it can achieve near ideal results. To validate the idea provided in this paper, a MetaService architecture which can provide an extensible and feasible implementation of the Sat Opt algorithm is designed. Experiment results show that the Sat Opt algorithm based on the user satisfaction level model can significantly improve the user satisfaction level compared with the related methods