English abstract
Grid computing has emerged as a new distributed computing paradigm. It offers a large scale of resource sharing and provides a variety of services.
A lot of research has been conducted on grid computing. One of the important research topics is load balancing in the grid. In this project, we conduct a performance evaluation of selected load balancing algorithms (Tree based algorithm and Ring based algorithm) in the grid. In this project, we consider users submitting independent tasks to grid system. We focus on how to distribute the load to the grid computing units in order to maximize the throughput and grid resource utilization. We developed both an analytical model based on queuing theory and a simulation model using NS2 a network simulation tool in order to derive performance characteristics of the selected load balancing algorithms used in the grid. results show that there We conduct simulation experiments which validate our analytical model. The results show that: the tree based algorithm (based on CPU speed) maximizes the throughput at high load. Its saturation level is very close to one.
