Attribute Partitioning in a Self-Adaptive Relational Database System.

摘要

One technique that is sometimes employed to enhance the performance of a database management system is known as attribute partitioning. This is the process of dividing the attributes of a file into subfiles that are stored separately. By storing together those attributes that are frequently requested together by transactions, and by separating those that are not, attribute partitioning can reduce the number of pages that must be transferred from secondary storage to primary memory in order to process a transaction. The goal of this work is to design mechanisms that can automatically select a near-optimal attribute partition of a file's attributes, based on the usage pattern of the file and on the characteristics of the data in the file. The approach taken to this problem is based on the use of a file design cost estimator and of heuristics to guide a search through the large space of possible partitions. The heuristics propose a small set of promising partitions to submit for detailed analysis. The estimator assigns a figure of merit to any proposed partition that reflects the cost that would be incurred in processing the transactions in the usage pattern if the file were partitioned in the proposed way. We have also conducted an extensive series of experiments with a variety of design heuristics; as a result, we have identified a heuristic that nearly always finds the optimal partition of a file.