Driver input selection for main-memory multi-way joins

Stream query processing has a particularly broad range of applications from sensor data processing and internet traffic analysis to runtime monitoring of stock market and server logs, and scientific simulations. This work focuses on multiway join queries over streamed data, which are processed with the help of a n-ary join. More specifically, we propose a novel main-memory variant of the influential M Join operator proposed in [35], which processes input data in batches with a view to improving the CPU efficiency, and explicitly controls the order of execution within each batch without being restricted by the time of input arrival, as current state-of-the-art solutions do. To this end, we also propose policies for selecting the execution order, and we show that our approach can yield important performance benefits.