Calibration and profile based synopses error estimation and synopses reconciliation

Yariv Matia, Yossi Matias

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

An important factor in the effective utilization of data synopses is the ability to have good a priori estimates on their expected query approximation errors. Such estimates are essential for the appropriate decisions regarding which synopses to build and how much space to allocate to them, which are also at the heart of the synopses reconciliation problem. We present a novel synopses error estimation method based on the construction of synopses-dependant error estimation functions. These functions are computed in a pre-processing stage using a calibration method. Sub-sequently, they are used to provide ad hoc error estimation w.r.t. given data sets and query workloads based only on their statistical profiles. We also present a novel approach to synopses reconciliation, using the error-estimation functions within synopses reconciliation algorithms, gaining significant efficiency improvements by lowering to a minimum and even avoiding interference to the operational databases. Our method enables the first practical solution for the dynamic synopses reconciliation problem.

Original languageEnglish
Title of host publication23rd International Conference on Data Engineering, ICDE 2007
PublisherIEEE Computer Society
Pages446-455
Number of pages10
ISBN (Print)1424408032, 9781424408030
DOIs
StatePublished - 2007
Event23rd International Conference on Data Engineering, ICDE 2007 - Istanbul, Turkey
Duration: 15 Apr 200720 Apr 2007

Publication series

NameProceedings - International Conference on Data Engineering
ISSN (Print)1084-4627

Conference

Conference23rd International Conference on Data Engineering, ICDE 2007
Country/TerritoryTurkey
CityIstanbul
Period15/04/0720/04/07

Fingerprint

Dive into the research topics of 'Calibration and profile based synopses error estimation and synopses reconciliation'. Together they form a unique fingerprint.

Cite this