Performance Comparison of Multi-Agent Grid Job Scheduler Organizations

J.S. Iyilade, G.A. Aderounmu (Nigeria), and M.O. Adigun (South Africa)


Computational organization theory, Grid Computing, Multi-agent, Scheduling


Multi-agent systems are currently been investigated by many researchers for scheduling job in grid computing systems. The organization of agents is such systems is however an important issue. In this paper, we present a performance comparison of two multi-agent grid job scheduler organizations – centralized and hierarchical organizations. We assume a multi-agent system where agents cooperate and distribute task using the Contract Net Protocol (CNP). A performance model was developed using Network bandwidth optimization, scheduling/mapping time, and Fault tolerance as performance parameters. The model was then simulated using Java programming language. Our simulation results show that the hierarchical organization optimizes network bandwidth, has a lower scheduling/mapping time, and responds better to fault when compared to the centralized organization.

Important Links:

Go Back