This is an outstanding text on transaction processing, covering both theoretical and practical aspects of the field. It will serve as an excellent textbook for graduate and advanced undergraduate courses. It will also be useful as a reference book for the researchers in the field.

The book consists of five parts. The first part contains motivation and background material. It discusses why transactions are important, and what is relevant. Also introduced in this part are two precise computational models for transactional servers. All discussions in the rest of the book are based on these two models.

The following two parts of the book are devoted to an in-depth presentation of concurrency control and recovery, which are the two basic components of transactional information systems. Part 2 is organized into eight chapters, discussing many different aspects of concurrency control. The discussion begins with an introduction of notions of correctness of data to be ensured when multiple transactions concurrently access shared data. A number of scheduling algorithms, used to satisfy the serializability notion, are described for both cases of single version and multiversion data items. Correctness notions and scheduling algorithms for object model (multilevel) transactions are also presented. An object model transaction is represented as a tree of operation invocations, rather than as flat sequences of simple read and write operations. Semantic approaches to concurrency control, concurrency control methods on search structures, and implementation and tuning techniques for concurrency control are the other issues discussed in the second part of the book.

Part 3 covers recovery methods used to protect data against system failures. The first issue discussed is transaction recovery, which is needed to implement transaction rollbacks while the server continues to process other transactions. Then, a notion of correctness, and algorithms, are described for the crash recovery of a data server. The algorithms are used to ensure that the data on stable storage are brought back to the most recent consistent state. Methods of recovering data from media failures, which correspond to damage to disk-resident data, are also presented. Besides data recovery, the issues in process recovery and message recovery are also addressed in this part.

Coordination of distributed transactions is the subject of Part 4. Concurrency control and recovery algorithms discussed in this part assume a distributed architecture comprising multiple application and data servers. Finally, Part 5 is devoted to a brief discussion of application perspectives and future trends.

Each chapter ends with lessons learned, exercises, and bibliographic notes. Examples are provided when appropriate. A Web site contains additional teaching materials and solutions to selected exercises.