SlideSide: A fast Incremental Stream Processing Algorithm for Multiple Queries

Aggregate window computations lie at the core of online analytics in both academic and industrial applications. To efficiently compute sliding windows, the state-of-the-art algorithms utilize incremental processing that avoids the recomputation of window results from scratch. In this paper, we propose a novel algorithm, called SlideSide, that extendsTwoStacks for multiple concurrent aggregate queries over the same data stream. Our approach uses different yet similar processing schemes for invertible and non-invertible functions and exhibits up to 2x better throughput compared to the state-of-the-art incremental techniques in a multi-query environment.

23rd International Conference on Extending Database Technology (EDBT)
Publication Year
Related Projects