Browse Books

Reviewer: Mukesh Singhal

Distributed database systems are very important because the availability of low-cost processors and advances in communication technology have made distributed data processing a viable means of information processing. This book is intended as a text for a graduate course in distributed database systems. Chapter 2 provides a nice overview of the concept of relational database systems, and almost every chapter starts with a discussion of the issues in centralized database systems and then proceeds to discuss distributed database systems. Nevertheless, readers are assumed to have some background in the basics of a centralized database system. A salient feature of the book is its broad coverage of topics and emphasis on fundamentals. Surprisingly, however, distributed deadlocks are not treated fully. A section of seven pages is not enough; a whole chapter should have been devoted to this topic. The strongest part of the book is the coverage of distributed query processing in chapters 7 through 9. The authors have done an excellent job of explaining the steps in query processing and query optimization. They nicely explain how different experimental and commercial systems perform query optimization. Chapter 11, on “Distributed Concurrency Control,” covers major concurrency control algorithms, but it underrepresents an important class of concurrency control, “concurrency control under data replication.” Data replication is touched on in chapter 12 under reliability issues. Chapter 12, on reliability, begins with basic concepts and definitions and then systematically explains various kinds of commit protocols. The authors explain in painstaking detail how two-phase commit deals with various kinds of failures (such as site failures and network partitioning). They discuss the types of failure in a distributed database system and give a detailed treatment of local recovery algorithms. This chapter does not discuss site recovery in distributed database systems (how a site merges with the rest of the system after a failure) and it lacks case studies. Also, the coverage of checkpointing is scanty, especially that of distributed checkpointing and its significance in recovery. Chapter 13 discusses operating system support for high-performance database systems. Conventional operating systems are not designed for database environments. Their functionality must be enhanced considerably to meet the demands of a transaction processing environment. These enhancements include facilities for distributed scheduling, interprocess communication and synchronization, distributed buffer and memory management, data caching, access control and protection, and support for transactions and locking. In this chapter, the authors lucidly explain how database operating systems can be designed to provide these enhancements to database systems. The treatment of buffer management issues is too long, however, especially since buffer management is not necessarily a distributed database issue but a centralized database issue. Chapter 14 covers one of the hottest topics in distributed database systems, multidatabase systems, in which several database systems are integrated so they look like a single, cohesive distributed database. The authors discuss three major issues in the design of multidatabase systems, namely integration of various databases, query processing over various databases, and transaction management. Although they deal with these issues very well, multidatabase serializability could have been explored more deeply. Multidatabase systems are still in their infancy and offer significant research opportunities. This chapter is an excellent introduction for someone who is planning to pursue such research. Chapter 15 covers several trendy topics such as distributed knowledge bases and object-oriented distributed databases. Though this coverage is well done, it will appeal to only a limited segment of readers. The authors use figures and examples lavishly to explain concepts. Each chapter contains “Bibliographic Notes” that provide pointers to sources for further reading. The chapters could use more exercises. Directory service should have been given more coverage. Performance evaluation of distributed database systems is not discussed, although much work has been done on the performance of concurrency control and query processing. This work is an excellent textbook for a graduate course in distributed database systems because missing and underemphasized topics can be supplemented by appropriate papers from the literature. The book is an excellent source of knowledge for researchers. For hands-on practitioners, however, the book may not be very useful because it merely provides many design alternatives without any comparison of their performance or much insight into which technique is suitable to what type of application.