Thin-client computing has become very popular in recent years, and this is indeed a surprise if one acknowledges the fact that it is very similar to the main-frame environments of decades past and if one considers the drive toward client-server computing in the 1990's. Advances in computing are usually thought of as a rupture with the past. The incorporation of thin-client computing is in this sense surprising. Clearly there are some advantages in using thin-client computing, at least for relatively small organizations. However, for large organizations it can be problematic, requiring meticulous attention to the administration of the servers and making sure that the client applications have been developed with thin-client environments in mind.

This book gives a helpful overview of thin-client computing using Microsoft Windows Server 2003 Terminal Services. It is clearly written for the administrator, but network design engineers or those who need to understand performance issues in thin-client environments still could read it profitably. Many businesses still insist on centrally managed computing, and 2003 Terminal Services was designed for those that do.

A major issue for large deployments of terminal servers that is glossed over in the book is the cooling requirements for server racks. For businesses that have thousands of clients, this entails a large data-center be constructed to house the server racks. Cooling becomes an expensive proposition, and in fact so critical that many small software companies have arisen in the last few years that offer algorithms, based on computational fluid dynamics, to optimize the air flow in these data centers.

The author correctly observes that testing in terminal server environments before deploying in production is extremely important, and a significant portion of the book is devoted to different ways of testing. He is also aware that those who do not actually do it usually consider testing a "necessary evil". In large terminal server environments it is even more crucial, and the collection of data, the creation of the statistics, the analysis and documentation become tedious but absolutely necessary.

Interestingly though the author does not feel that simulation modeling is of too much value when it comes to understanding the behavior of traffic in terminal server environments. He claims, but does not back this up with explicit calculations, that it is more cost-effective to purchase too many servers than to invest in model development. Modeling, whether discrete simulation modeling or based on relevant mathematics can be expensive sometimes there is no doubt. However, it allows insight into the behavior of devices in production environments, and this insight is frequently superior to what can be obtained using testing or educated guesswork. Interestingly however, the major vendors of discrete simulation software have not expressed any interest in the RDP protocol, this due in part to the difficulty in obtaining the source code for this protocol. Therefore, the development of a simulation model that emulates the data traffic of a user over RDP will have to be done essentially from scratch. The complications due to user behavior, such as mouse events, typing rates, and the caching and queuing mechanisms at the client can be incorporated in such models.

It is clear from reading the book that the author has been involved in large-scale terminal server deployments. His advice is therefore valuable to those readers who are faced with the same. And there are no doubt many readers who, having stood up terminal server environments, will back up the author's statements with gusto. Issues with printing, network latency, caching at the client, and especially load balancing and how it meshes with Session Directory are things that network engineers and administrators are faced with frequently in terminal server environments. The author gives detailed steps on how to avoid or at least handle these issues, this being done through careful planning and foresight, with intuition and `gut feelings' to be completely avoided.