An Enhanced Tuple Routing Strategy for Adaptive Processing of Continuous Queries

Abstract

Work on efficient processing of long running queries on data flows has attracted much attention nowadays in modern data base management systems. The query optimizer within every query processing engine uses statistical properties to select an efficient execution plan. Since many application programs work with data flows, in a query processing engine which supports long running queries on data resources, the collection of statistical properties for data resources and servers, such as input data flow rate and available computational resources that change during execution, is quite difficult. The use of traditional query optimizers for the processing of these types of queries is thus inappropriate, since the execution map of a query must be capable to cope with these changes. So, adaptive processing of long running queries on data flows must be used instead In routing-based adaptive processing of queries, optimization is performed during execution and tuple routing is used as an adaptive optimization technique, there is no explicit execution map anymore and each tuple is routed uniquely according to a defined routing strategy. We present a new tuple routing strategy to improve the performance of adaptive processing of continuous queries. We use a time window to measure changes in data streams, in addition to specific properties such as operator cost, operator selectivity and operator message queue length. Experimental results show favorable improvements with respect to existing strategies.