The load your users place on your Windows 2000 servers may increase approximately linearly over time, but that's no guarantee that the servers' performance will degrade smoothly and predictably. Rather--and this is the crux of Windows 2000 Performance Guide--a modest increase in workload can often cause a significant, even catastrophic, decrease in overall performance. The reason: Servers are complex systems, with each of their parts dependent upon many others. Your job in optimizing Windows 2000 machines is to spot the critical thresholds (preferably in advance) and adjust your systems to stay clear of them. Mark Friedman and Odysseas Pentakalos have done a considerable amount of empirical research into the behavior of all major Windows 2000 subsystems (and make frequent, detailed reference to the research of others) and present their findings here. Their approach is somewhat academic (you can't accurately describe performance without some calculations and statistics, as well as some theoretical discussion of operating system design), but there's no question that this volume fits into the "blue" series of system administration books. Software engineers--particularly those engaged in designing highly scalable applications for Windows 2000--will get a lot from this title, as well.
For starters, the authors go into depth on what the traces available in Performance Monitor mean--valuable stuff for a system administrator who's unclear on what "Context switches / sec" and "Thread % User Time" (to cite one example) have to say about how efficiently available resources are being used. Most of the instruction comes in the form of laboratory narratives that describe symptoms, systematic observations (i.e., how Perfmon was used for diagnosis), and corrective adjustments. In most cases, the authors explain the relative merits of better hardware, parallel hardware, application tuning, and other alternative solutions, but leave it up to readers to best resolve their own systems' particular troubles. --David Wall
Topics covered: The subsystems of Microsoft Windows 2000, how they interact, and how they affect overall system performance under different applications. Performance Monitor is dealt with in depth, as are the performance characteristics of CPUs (single and parallel), memory and paging, disk access and caching, network access, and Internet services. Threads and their priorities are explained in easily understood detail.
It is characteristic of most computer systems that they do not degrade gradually. The painful reality is that performance is acceptable day after day, until quite suddenly it all falls apart. When this happens, the administrator needs to be prepared to help the organization get through the crisis. Computer applications are growing ever more intelligent and easy to use. One of the by-products of making applications easier to use is that they usually also require more resources to run. And wherever productivity is a central factor in the decisions you make, performance considerations loom large and continue to play an important role in system management. Are you wondering, for example, if more expensive equipment would give better performance? The answer is often yes, but not always. This book will show you why it is important to understand the performance characteristics of the hardware and of the workload, and how they match up against each other. Windows 2000 Performance Guide takes you through problem solving techniques like measurement methodology, workload characterization, benchmarking, decomposition techniques, and analytic queuing models.