Evaluating Window Joins over Unbounded Streams

Speaker: Kin Kong LOO

Abstract

Recently, problems related to unbounded, continuous input streams have attracted much interest. In a research work published in ICDE '03, Kang et al investigate algorithms for evaluating sliding window joins over pairs of unbounded streams. A unit-time-basis cost model is introduced to analyze the expected performance of these algorithms. Using this cost model, strategies are proposed for maximizing the efficiency of processing joins when one stream is much faster than the other, and for maximizing the number of tuples in the join result when computing resources and/or memory are inadequet for getting the exact join result.