Dynamic scheduling algorithm for real-time applications in grid environment

Advances in networking infrastructure have led to the development of a new type of "computational grid" infrastructure that provides predictable, consistent and uniform access to geographically distributed resources such as computers, data repositories, scientific instruments, and advanced display devices. Such Grid environments are being used to construct sophisticated, performance-sensitive applications in such areas as supercomputer-enhanced instruments, desktop supercomputing, tele-immersive environments, and distributed super computing. Such applications designed to execute on "computational grids" frequently require dynamic scheduling of multiple resources in order to meet performance requirements. Motivated by these concerns, we have developed a general scheduling algorithm in Grid environment. In this paper, we propose a three level dynamic scheduling method. A set of thresholds and information list are used to collect the information of the grid. In first level, as task arrive in the one node of the system. the scheduler uses the information about the task and the state of node, and attempts to guarantee the new task by the execution of this node. If this task can be scheduled on this node and the time constraints can be met, we execute this task on this node. If this task cannot scheduled, the second level scheduling is started., we use the information list of the cluster that is connected closely by the node to find whether this Cluster can accept this new task and meet its time constraint. If the cluster can accept this new task, we transfer this new task to one underloaded node of this cluster and executes this new task on this underloaded nodded. If new task cannot be accepted by the cluster closely connected by the node, the third level scheduling is started., we use the information list to find a remote cluster in the grid to accept this task and execute it on the remote cluster in the grid.