关键词: 网格计算, 可视化调度, 综合效用率, 后调度

Abstract: To improve scheduling performance and better meet users requirements in grid applications oriented to engineering computation, a user-centered scheduling strategy was proposed. Within this strategy, three performance metrics were modeled, i.e. Degree of Credit (DC), Degree of Guarantee (DG) and Aggregate Utility Ratio (AUR)… Among these metrics, AUR, as a composite Quality-of-Service (QoS), was composed of four performance factors with their individual weights. To deal with different types of grid users, automatic scheduling, manual scheduling and associated steering-enabled visual interfaces were designed respectively. Based on an overall scheduling framework, the visualization of all performance metrics has facilitated users interaction with QoS-driven scheduling mechanism. For those scheduled tasks which could not obtain expected QoS, a specific post-scheduling mechanism was applied to cope with all exceptions. Some tests were performed on a grid test bed of Vgrid. These experiments indicated that the visual scheduling was suitable for scientific and engineering oriented grid computing in practice.

Key words: grid computing, visual scheduling, aggregate utility ratio, post-scheduling