Evolutionary techniques for updating query cost models in a dynamic multidatabase environment

Abstract

Deriving local cost models for query optimization in a dynamic multidatabase system (MDBS) is a challenging issue. In this paper, we study how to evolve a query cost model to capture a slowly-changing dynamic MDBS environment so that the cost model is kept up-to-date all the time. Two novel evolutionary techniques, i.e., the shifting method and the block-moving method, are proposed. The former updates a cost model by taking up-to-date information from a new sample query into consideration at each step, while the latter considers a block (batch) of new sample queries at each step. The relevant issues, including derivation of recurrence updating formulas, development of efficient algorithms, analysis and comparison of complexities, and design of an integrated scheme to apply the two methods adaptively, are studied. Our theoretical and experimental results demonstrate that the proposed techniques are quite promising in maintaining accurate cost models efficiently for a slowly changing dynamic MDBS environment. Besides the application to MDBSs, the proposed techniques can also be applied to the automatic maintenance of cost models in self-managing database systems.

Towards Adaptive Costing of Database Access Methods

“…By applying this procedure repeatedly, so that the current model is based on recent observations, we can produce cost models that adapt automatically to changes in the database system configuration or operating point. Similar approaches have been proposed for modeling the costs of complex XML query operators [14], for costing UDF executions [7], and for costing remote sub-queries in federated database systems [11]. However, we argue that the problem of cost estimation for database access methods presents some unique challenges.…”

“…Examples include estimating the costs of user-defined functions [7], estimating the costs of remote sub-queries in federated database systems [11], and estimating the costs of complex XML query operators [14]. The so-called "parade-of-runs" approach [1] relies on a synthetic calibration workload to produce a training data set that covers the entire input parameter space.…”

A Bi-objective Cost Model for Database Queries in a Multi-cloud Environment

“…[21] provides more detailed cost functions but follows the same approach as in [11]. More sophisticated approaches, such as [24,16], perform more efficient and dynamic cost function calibration but still suffer from the main limitations mentioned above.…”

Efficiently Updating Cost Repository Values for Query Optimization on Web Data Sources in a Mediator/Wrapper Environment

“…Work by Rahal et al [12], focused on multidatabases, propose a couple of approaches so that a cost model is able to get adjusted when the costs change; their first approach is interesting from a mediator point of view (the second one requires that changes are not abrupt so there is no need of a new model for each value measurement, which, in the case of web data sources characterization, can not be confirmed at all), even though cost information is obtained directly from the repositories (currently imposible to achieve in a web environment). Rahal defines a cost model as the independently-weighted sum of the characterization parameters, and describes a series of techniques to adjust them when the database behaviour changes.…”