An Adaptive Process Allocation Scheme in Grid Environment

The Grid is an interconnected set of distributed compute servers. An application running on the Grid consists of processes, which can be executed in parallel or in a sequential manner. An application can specify application level and network level Quality of Service parameters including number of processors, memory, special software, network bandwidth, delay, jitter, packet loss, etc. We investigate the question: Which processes are allocated to which compute servers that collectively satisfy the application's resource requirements and optimize performance and cost parameters. We describe a protocol to identify those compute servers that can execute the application with minimal cost and provide the required application level and network level Quality of Service