This is a book about database management systems (DBMSs) that manipulate objects. These systems have been develped to support new kinds of applications--applications whose needs are quite different from those of traditional business database applications. These applications include computer-aided design (CAD), computer-aided software engineering (CASE), computer-aided manufacturing (CAM), office automation, scientific applications, telecommunications, expert systems, and other applications with complex and interrelated objects and procedural data. The new database systems also open possibilities for traditional business applications.
In this book, we study a wide variety of DBMSs advanced since the development of relational systems in the early 1980s. We examine new applications and the characteristics of their data, in order to understand the shortcomings of current DBMSs and the need for object data management systems (ODMSs) for these more sophisticated uses. The ODMSs we study include object-oriented DBMSs, extended relational DBMSs, functional-semantic DBMSs, DBMS generators, and simple object managers. We examine specific commercial products and research prototypes for object data management and compare the merits of different approaches.
Some readers might call this a book about object-oriented database systems, inasmuch as "object-oriented" has been used by one author or another to characterize nearly all the approaches we cover. However, in order to avoid confusion among divergent approaches, we use the term "object data management." Object-oriented database programming languages Bancilhon and Buneman 1990 are the DBMSs most precisely called "object-oriented." The primary focus of the book is on these systems and on extended relational database systems, because these approaches have been the two most widely accepted in the database research and development communities.
This book is aimed at an engineer, programmer, technical manager, or student interested in recent advances in database systems. In addition to introducing basic concepts and approaches to object data management, it covers important factors that the reader might overlook when comparing and using the new database systems. The area of object data management is new and is not well understood; this book sorts through conflicting claims for these new DBMSs. There are substantial differences in the performance and functionality among existing products and approaches.
The book is also appropriate for the reader involved in designing and impelementing a DBMS. It provides a review of new object data management technology, with references to more detailed coverage of specific implementation issues. In contrast to most other sources currently available, the book provides a comparison of different object approaches from a unified, balanced perspective, and a study of database architecture requirements on the basis of application requirements, rather than the most popular research areas.
This book complements, and may be supplemented by, existing collections of articles Zdonik and Maier 1989, Kim and Lochovsky 1988, Cardenas and McLeod 1990, Bancilhon and Buneman 1990.
We assume that the reader has basic familiarity with computing and programming concepts. Background in traditional DBMSs, particularly in relational systems, will also be important; however, we review conventional DBMSs, and references are provided for more background. This book would be a good text for a second course in database systems, to follow an introductory text such as Date 1990, Ullman 1989, Elmasri and Navathe 1990. Alternatively, the book could be used for a first course by instructors wanting to put special emphasis on next-generation database systems.
The reader needs no background in applications such as CAD, CASE, and office automation. In fact, we begin by covering these, because it is important to understand applications before proceeding honestly with a study of DBMSs to satisfy the application's needs--a fact often forgotten. A study of any one of these applications could fill entire volumes in itself McClure 1989, Rubin 1987, Groover and Zimmers 1984, however, relatively simple examples will illustrate the needs for new database functionality.
The need for object data management sprang from advances in software technology for the applications we study in this book, particularly CASE and CAD, enabled by the widespresd introduction of engineering workstations and networking in the past decade. These applications have become more powerful and better integrated. Application designers and users naturally are seeking the next step: a DBMS that can provide a higher level of integration, acting as a common repository for objects shared by many applications and users.
Many engineering and office systems are already built around custom "database systems," with data structures that can be written out to disk files for interchange with other software. These custom database systems are not true DBMSs, however: true DBMSs are application-independent and provide many more features. Unfortunately, traditional DBMS products have not provided the performance and features these applications need. ODMSs were developed to meet those needs.
The business applications market has undergone a major revolution, moving from application-specific index-based file storage to general DBMSs, and more recently to relational DBMSs Codd 1970, Date 1990. However, it is difficult to modify traditional DBMSs, including relational DBMSs, to satisfy the performance and functionality requirements of the applications we study in the book. As we shall see, quite different architectural tradeoffs are necessary to support their needs.
This book addresses a new DBMS revolution, as CASE, CAD, office automation, and other applications move to the more powerful and flexible platform of object data management. ODMSs are also likely to be used in traditional business applications, as many ofthe new capabilities are important there as well.
Enhancements in Revised Edition
A number of corrections and improvements have been made to this book since it was published in 1991. This revised edition represents a substantial update to bring the book up-to-date with the latest technology and events in this fast-moving field, and to improve the presentation based on feedback from readers and instructors using the book over the last few years.
A new chapter (Chapter 6) has been added to cover standards work in OMG, ODMG, and the ANSI SQL committee. As noted in the first edition, standards were an important missing link in the future of object data management. The recent completion of the ODMG-93 standard as well as progress on a SQL3 standard are very important to new database technology.
The Appendix covering products and prototypes has been expanded and updated. The original edition covered fewer systems in detail. Since many readers are using this book to make a decision between alternative products, or are using a particular product or publicly-available research system for hands-on experience in conjunction with this book in an advanced database course, a more extensive coverage of the major systems has been added to the Appendix. The existing sections of the Appendix have been updated, and expanded in some cases.
Corrections and improvements have been made throughout the book. The notation used in the data schema figures has been improved. New figures and examples have been added. The bibliography has been expanded. Corrections and clarifications have been made in all of the chapters. MAterial has been added to cover new work in a number of areas.
In contrast to when this book first appeared, there are now many textbooks available on the topic of object-oriented databases; however, this book generally represents a more thorough and balanced coverage of new database technology. It discusses a much wider range of approaches, giving attention to extended relational as well as object-oriented DBMSs. It also provides deeper insight into the implementation and architecture of these systems; because the new systems differ more in architecture and performance than in features and data models, this insight is essential to both the student and industrial user of these systems.
Reading Instructions
This book is divided into three main parts. The first three chapters comprise background material that provides an overview of object data management, examines applications that motivate this new technology, and reviews traditional database systems. Chapters 4 and 5 provide a core study of object data management. Chapters 6,7, and 8 provide analysis and comparison of approaches and standards for object data management, with predictions for the future. The seven chapters are organized as follows:
Chapter 1, Introduction, covers basic data-modeling concepts that will
be important throughout the book, and explains the history and
motivation for object data management.
Chapter 2, Advanced Database Applications, covers new database
applications in detail; these applications will serve as examples
for the remainder of the book. The examples will be useful to
illustrate specific database functionality and performance we
need for object data management. At the end of Chapter 2, we
summarize the database needs of the new applications.
Chapter 3, Traditional Database Systems, reviews data management, with
a particular focus on relational systems. This chapter is included
not simply for historical perspective; much of object data management is
based on extensions to the traditional technology. Readers with a
background in database systems may skip this chapter, or may use
it as a review; for other readers, this chapter provides an overview,
with pointers to reading in specific topics.
Chapters 4 and 5 cover object data management. Chapter 4, Object
Data Management Concepts, covers functional issues, such as complex
objects, type hierarchies, procedures, and version management. Chapter
5, Implementation Issues, covers physical implmentation tradeoffs, the
programming-language interface, query processing, remote and
distributed access, and overall system performance.
Chapter 6, Future Database Standards, covers work in a number of
groups that are likely to shape future standards for object data
management.
Chapter 7, Goals for Object Data Management, provides a basis to
compare ODMSs, deriving a set of basic properties that new database
systems should satisfy.
Chapter 8, Conclusions, evaluates the different approaches to object
data management. It also covers future directions for research on
ODMSs, and reviews what we have learned in the book.
In the Appendix, Products and Prototypes, we study specific commercial products and research prototypes to illustrate further the approaches to object data management discussed in the book: extensions to relational DBMSs, object-oriented DBMSs, functional and semantic model DBMSs, database generators for building tailored systems, and simple object managers based on file system extensions.
In addition to an Index, the book includes an Annotated Bibliography of reference materials covering seminal papers in object data management, new object data management systems, and important background material in programming languages and traditional database technology. A smaller System Bibliography indexed by system name is included at the end of the Appendix, for convenience in finding documentation for each system. The System Bibliography also serves as a glossary of terms used to categorize ODMSs, with a categorization of all the major ODMS products and prototypes.
For use in a course on advanced database systems, the book can be covered in approximately four equal portions taking about 8 instructional hours each: (1) introduction and motivation in Chapters 1 and 2, (2) object data management concepts in Chapter 4, (3) object data management implementation in Chapter 5, and (4) future work in object data management in Chapters 6 through 8. Chapter 3, on traditional database systems, and the Appendix, detailing specific systems, are optional depending on course background and goals.
For the reader using this book to evaluate object data management systems for use in a specific application, Chapters 1, 4, 7, and 8 are essential reading. Chapters 2 and 3 are optional, depending on background in applications and traditional database technology; Chapters 5 and 6 should be useful as a reference. The Appendix should be useful for readers examining specific products and approaches.
Obtaining Product Information
Readers planning to purchase one of the ODMSs described in this book will probably want to consult additional sources before making a large investment. Best efforts have been made to assure that the systems described in the text and in the Appendix are accurate and up-to-date, and this book probably covers more systems in more detail than any other; however, it is not feasible to do detailed comparisons of these systems in a book, nor is it possible to update the information as quickly as the products change.
To assist readers with this problem an independent consulting firm, Barry and Associates, has prepared a quite comprehensie, regularly-revised report on ODMS products. This report is the best source to my knowledge for making a product selection, and it was prepared wit hmy consultation. See the bibliography Barry and Associates 1994 for information on obtaining this report.
Corrections and Suggestions
Compiling a book categorizing and analyzing half a dozen different approaches to database systems, and referencing hundreds of technical papers that have never been brought together in a single discussion of future directions in database systems, has been a substantial task. Many of the areas covered are not well understood, or are subject to substantial disagreement in the database research community. Despite these problems, I have attempted to provide an understanding of the issues and to take a position on most of them. I hope I have elucidated more often than I have confused in so doing. My apologies to any authors whose work or positions I have represented poorly. Since the study of this subject is so vulnerable to errors, disagreements, omissions and confusion, your input is solicited for future editions; the area is changing rapidly, so it is likely that a new edition will be needed soon. Comments, corrections, and constructive suggestions should be sent to Addison-Wesley, Reading, Massachusetts, 01867 or by electronic mail to:
odmbook@eng.sun
Acknowledgments
My wife, Susan, deserves special recognition for her substantial and flexible support of this book and of my career. I also thank my parents for their encouragement in my academic endeavors, and my sons, Eric, Aaron, and Elliott, for sharing me with the work I bring home.
Leon Guzenda, David Jordan, Mohammed Ketabchi, Fred Lochovsky, Bill Paseman, and Mike Robson deserve special thanks for the substantial effort they invested i...
