aqp-tutorial-description

AdaptiveQueryProcessing: Why,How,When,WhatNext?Amol Deshpande Zachary Ives Vijayshankar RamanUniversity of Maryland University of Pennsylvania IBM AlmadenABSTRACT 2. MOTIVATIONSFORAQPDeclarative queries are a central value proposition of theAdaptive query processing has been the subject of a greatrelational model, letting the users specify only what re-deal of recent work, particularly in emerging data man-sults they want without having to worry about the strat-agement environments such as data integration and dataegy (plan) used to access and combine the data. Findingstreams. We provide an overview of the work in this area,the best plan (query optimization) was addressed in evenidentifyingitscommonthemes,layingoutthespaceofquerythe first RDBMSs — most successfully by Selinger et al.’splans, and discussing open research problems. We discussdynamic programming algorithm in System R. System Rwhy adaptive query processingis needed, how it is being im-divided query processing into separate optimization and ex-plemented, where it is most appropriately used, and finally,ecution stages and used cost-based enumeration of possi-what next, i.e., open research problems.ble query plans. Over time, this optimization approach hasbeen improved (exploring more exhaustive plans, using his-1. INTRODUCTION tograms, adding cross-block query rewrites), but the basicSystem R architecture lives on in most query processors.In recent years, there has been increasing use of ...