Nowadays, network architectures are a critical part of any parallel or distributed system. An incorrect choice of the network topology connecting different processing elements will surely decrease system performance. This book by Tutsch studies the topological characteristics of different network architectures, and provides a detailed analysis of network performance based on mathematical methods, such as Markov chains and Petri nets. The book is not intended for students or practitioners that want to know the internals of existing network architectures. For these readers, more general and well-established books [1] will be more appropriate. Instead, this book is primarily intended for researchers interested in network performance modeling. After a brief introduction and motivation, chapter 2 discusses the main characteristics of existent network architectures, such as switching techniques and wired and wireless network architectures. Some comments about the intrinsics of network-on-chip (NoC) architectures and network reconfiguration are also given. This chapter provides a good overview of network topologies.
Chapter 3 is devoted to the performance evaluation of network architectures. Both simulation and mathematical methods are presented, the latter including Markov chains and Petri nets. These modeling techniques are carefully discussed, together with many additional references. As pointed out by the author, these methods, while accurate, quickly become too complex, consuming much development time and needing many computer resources to be useful. Therefore, chapter 4 discusses some guidelines for model development and complexity reduction.
The last part of the book shows how to apply the concepts described in the previous chapters to different network architectures. Chapter 5 deals with a cellular network, a simpler example in which using a model based on Petri nets is enough. Chapter 6 studies a multistage interconnecting network, a more complex network in which both a simulation stage and a mathematical model are used. The complexity of the problem becomes apparent in this chapter, and the reader quickly understands that complexity reduction techniques are needed to overcome the high computation times needed by a simulation model.
The book is very carefully edited, and the number and type of references provided show that the author has researched this topic considerably. However, instead of just giving eight or ten references for Markov chains or Petri nets, a more detailed classification of all these references would have been more useful. In summary, this book is an excellent starting point for readers that want to apply simulation techniques and mathematical methods for measuring network performance.