6.1 Introduction

Most large organizations have collected a considerable amount of data and have invested heavily in systems and applications to manage and access that data. Even within a single organization, these legacy data management systems typically vary widely, ranging from simple text files with little or no support for queries or other database functionality to complex database management systems with sophisticated query engines. In many cases, specialized indexing technology and query engines have been developed to facilitate efficient searching of particular kinds of data, for example, for finding compounds with a certain substructure in chemical databases, for finding overlapping regions in a geographic database, or for finding images with similar color or texture in an image archive. It is increasingly clear that powerful applications can be created by combining information stored in these historically separate data sources. For example, a medical system that integrates patient histories, EKG readings, lab results and MRI scans will greatly reduce the amount of time required for a doctor to retrieve and compare these pieces of information before making a diagnosis. Likewise, a pharmaceutical application that can combine a chemical compound-similarity search with the results of biological assays will be a powerful tool for discovering new drugs.

There are several approaches to providing such an integrated view of heterogeneous data. One is to move the data en masse to a new integrated database system that is tailored to provide a unified...