Aggregate Queries over Data Streams Based on Multi-Bloom Filters

This paper targets at aggregate queries over historical data in data stream within time intervals,and proposes a novel storage model called Multi-Bloom Filters(MBF) based on Bloom Filters(BF).It uses a global bit vector to realize high efficiency of insertion and query,combines dynamically allocated local counter vectors to store historical data over different time intervals into these counter vectors.Therefore,MBF efficiently supports to store and query historical data over data stream using multiple levels of time granularities.A method of compressing space of MBF is given under the condition that the time span is very large.Analysis shows that MBF has great flexibility and supports approximate aggregate queries over historical data within time intervals,where optimal parameters of MBF are provided in term of query accuracy.