Establishing Dynamic Cost Models for Multidatabase Systems

在多数据库系统中进行全局查询优化的一个关键挑战是，由于局部自治，某些所需的局部信息(如局部代价模型)可能在全局级别不可用。最近的文献中提出了一些应对这一挑战的技术。但是，它们仅适用于静态系统环境。本研究项目探索建立多数据库系统动态成本模型的理论和方法。先前关于用于在静态系统环境中建立成本模型的查询采样方法的工作被扩展到动态系统环境。本项目采用的方法包括在成本模型中引入定性变量的定性方法和自适应方法，该方法自适应地将从执行用户查询中观察到的动态成本信息纳入成本模型。对所有方法进行了评估和比较，以确定最有前景的方法来建立多数据库系统的动态成本模型。并对基于动态代价模型的查询优化进行了理论和实验研究。该项目将为设计和开发高效的多数据库系统提供理论、算法和实现技术。

A crucial challenge for global query optimization in a multidatabase system is that some required local information such as local cost models may not be available at the global level due to local autonomy. A number of techniques to tackle this challenge have been suggested in the literature recently. However, they are suitable only for a static system environment. This research project explores theory and methods to establish dynamic cost models for a multidatabase system. Previous work on a query sampling method for establishing cost models in a static system environment is extended to a dynamic system environment. The methodologies adopted in this project include a qualitative approach, which introduces qualitative variables into cost models, and an adaptive approach, which adaptively incorporates dynamic cost information observed from execution of user queries into cost models. All approaches are evaluated and compared in order to determine the most promising one to establish dynamic cost models for multidatabase systems. Theoretical and experimental studies on query optimization that takes advantage of dynamic cost models are also conducted. The project will provide theory, algorithms, and implementation techniques for the design and development of an efficient multidatabase system.