Indexing the Results of Sliding Window Queries ∗

A popular method of bounding the memory requirements of queries over data streamsis to use a sliding window, where old data are continuously removed as new data arrive.One problem that has not been addressed previously concerns indexing the results of slidingwindow queries. This is a noteworthy problem because data stream systems often materializecommon sub-expressions or final results of similar queries, therefore it is important to allowthe queries efficient access into a relevant subset of the materialized result. In this paper, wedesign and evaluate indexing methods that take advantage of the temporal order in whichold answers expire for efficient maintenance and querying of the results. Our techniquesallow sharing of indexed results among similar queries, adapt to fluctuating stream arrivalrates, and are experimentally shown to perform updates over twice as fast as the existingsliding window indices.