Browse Books

Reviewer: Burkhard Englert

Despite the attention distributed computing has received in recent years, designing and analyzing a distributed system is still a very difficult task. Because of inherent asynchrony, limited local knowledge, and partial failures, it is challenging to gain a good understanding of a proposed system. On the other hand, the multiplicity and variety of distributed systems and their widespread differences make it desirable to describe a more general distributed computing universe where one abstracts away the system-specific details. In this exciting new textbook, Santoro follows this approach. At first glance, this book appears to be one of the standard textbooks on distributed algorithms, but a closer look reveals several features that distinguish it from other books with similar intentions. Namely, the book, using solely the message-passing paradigm, is based on a fully reactive computational model. This means that any entity in the system responds only to external events. These external events may be the arrival of a message, the ringing of the alarm clock, or a so-called spontaneous impulse. I agree with the author that this approach makes learning easier since it simplifies explanations and allows instructors to present the material in a natural way. As an added benefit, all protocols presented in the book and all answers to exercise problems are immediately programmable; as a result, it is possible to integrate experimental analysis into the coursework. The book is suitable for advanced undergraduate students, but probably will find its true audience among graduate students and other researchers who want to learn how to solve distributed computing problems. It begins in chapter 1 with an introduction to the distributed computing environment, where the author defines the model he uses throughout the book. In the first few pages, the reader can sense that this book is not simply a collection of research papers written in the area over the last few decades. The book is a pleasure to read right from the beginning. The style of presentation makes it obvious that students are the intended target audience. Because of the clarity of its presentation and its casual, yet precise and rigorous style, there is a good chance that students who need to learn this material will actually want to read the book. In the second chapter, the author moves on to some basic problems and protocols. This is followed by a long and thorough chapter on the problem of symmetry breaking (election) in distributed systems. The author discusses the problem in many different contexts. The fourth chapter deals with the efficiency of distributed communication, namely message routing and shortest paths. The next chapter discusses problems that arise in the context of distributed sets of data. In the first five chapters, all computation is assumed to take place in an asynchronous distributed system. In the sixth chapter, problems that arise in synchronous systems are introduced. Chapter 7 is another long and thorough chapter dealing with the issue of computing in the presence of faults. In chapter 8, the author addresses problems of detecting stable properties, such as deadlock or termination. In chapter 9, the book’s last, problems that require continuous computation, such as computing logical time and distributed mutual exclusion, are briefly discussed. This appears to be the weakest chapter in the book. Readers may be put off by the brief discussion that these important problems receive. On the other hand, this treatment is consistent with the philosophy of the rest of the book, which aims to teach readers how to solve distributed computing problems and not be an encyclopedia of research results. As such, the book is very well thought-out and implemented. The author uses his many years of teaching experience to present this difficult material in a way that makes the reader want to solve distributed computing problems. I believe that if this book receives the attention it deserves, it will become a classic textbook in distributed computing theory. It has the potential to educate, motivate, and enable current and future students and researchers to effectively solve distributed computing problems. Extending its reach, its style of presentation and language are also very appropriate for today’s system programmers or communication software engineers and developers—it is not just another book about the principles of distributed computation. Online Computing Reviews Service