【摘要】 在网格计算环境下,许多任务的执行经常需要同时协同分配多个agent以满足性能需求。提出了一种新的基于multi-agent编组机制的agent协同分配方法。首先,对传统编组算法和multi-agent编组算法进行了分析比较。然后,设计了校园虚拟实验网格调度系统及其任务分配算法,实验结果表明:当CDR=0.2时,算法与传统算法带来最好的目标价值,在某些条件下,传统算法略微胜过该算法,当CDR增大时,该算法比传统算法执行得更好。更多还原

【Abstract】 Many task designed to execute on grid computing frequently request the simultaneous co-allocation of multiple task in order to meet performance requirements.In this paper,a new approach via multi-agent coalition mechanism is presented for agent co-allocation.Firstly,the traditional coalition algorithm and ours are analyzed and compared.Then,a campus virtual experimental grid scheduling system and its mission allocation algorithm are designed.The experiment result indicated that the same tendency: when CDR=0.2,the traditional algorithm and ours brought the best objective values,and the traditional scheme slightly outperformed ours under some conditions,and as CDR became larger,our scheme performed better than the traditional one.更多还原