A multicore parallelization of continuous skyline queries on data streams

Skyline queries are preference queries frequently used in multi-criteria decision making to retrieve interesting points from large datasets. They return the points whose attribute vector is not dominated by any other point. Over the last years, sequential and parallel implementations over static datasets have been proposed for multiprocessors and clusters. Recently, skyline queries have been computed over continuous data streams according to sliding window models. Although sequential algorithms have been proposed and analyzed in the past, few works targeting modern parallel architectures exist. This paper contributes to the literature by proposing a parallel implementation for window-based skylines targeting multicores. We describe our parallelization by focusing on the cooperation between parallel functionalities, optimizations of the reduce phase, and load-balancing strategies. Finally, we show experiments with different point distributions, arrival rates and window lengths.