Modelling Resource Utilization in Pipelined Query Execution

Database parallelism offers the potential of efficient query execution, but increases the complexity of optimization, as the impact of the workload of processors and network on competing and communicating processes must be considered. We introduce a query execution model that incorporates the effects of the system workload on pipelined query execution. Our model is general enough to cover bushy and pipelined parallelism for shared-nothing and shared-disk architectures.