Skip to Main Content
The problem of allocating processes of a Grid application to compute servers can be formulated as a distributed search problem. The goal is to allocate an applications processes to those compute servers that can provide the required Quality of Service and can execute the processes in a cost-efficient way. We present a protocol to identify compute servers that can execute the application with minimal cost and provide the required Quality of Service for the application. Relying on a heuristic search algorithm, the protocol makes use of statistical data of the compute servers' efficiency in past process executions.