Prefilter: predicate pushdown at streaming speeds

  • Authors:
  • Lukasz Golab;Theodore Johnson;Oliver Spatscheck

  • Affiliations:
  • AT&T Labs - Research;AT&T Labs - Research;AT&T Labs - Research

  • Venue:
  • SSPS '08 Proceedings of the 2nd international workshop on Scalable stream processing system
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents the prefilter: a predicate pushdown framework for a Data Stream Management System (DSMS). Though early predicate evaluation is a well-known query optimization strategy, novel problems arise in a high-performance DSMS. In particular, (i) query invocation costs are high as compared to the cost of evaluating simple predicates that are often used in high-speed stream analysis; (ii) selectivity estimates may become inaccurate over time; and (iii) multiple queries, possibly containing common subexpressions, must be processed continuously. The prefilter addresses these issues by constructing appropriate predicates for early evaluation as soon as new data arrive and before any queries are invoked. It also compresses the bit vector representing the outcomes of pushed-down predicates over newly arrived tuples, and uses the compressed bitmap to efficiently check which queries do not have to be invoked. Using a set of network monitoring queries, we show that the performance of the Gigascope DSMS is significantly improved by the prefilter.