Querying sliding windows over on-line data streams

Abstract

A data stream is a real-time, continuous, ordered sequence of items generated by sources such as sensor networks, Internet traffic flow, credit card transaction logs, and on-line financial tickers. Processing continuous queries over data streams introduces a number of research problems, one of which concerns evaluating queries over sliding windows defined on the inputs. In this paper, we describe our research on sliding window query processing, with an emphasis on query models and algebras, physical and logical optimization, efficient processing of multiple windowed queries, and generating approximate answers. We outline previous work in streaming query processing and sliding window algorithms, summarize our contributions to date, and identify directions for future work. © Springer-Verlag 2004.