Capacity Planning for Internet Services

The demands placed on a server farm by e-commerce and other Internet services differ from those imposed by longer-studied and better-understood processing tasks, largely because the potential demand for publicly accessible Internet services is practically infinite. Capacity Planning for Internet Services takes a careful look at traditional mainframe capacity planning techniques in light of the Internet's characteristics, evaluating their effectiveness in the new environment. The book offers new tools, formulas, and tables for use in estimating server requirements (they focus on the Sun Microsystems product line) for Internet work.

Ideal for system administrators, design engineers, database administrators, and anyone else involved in planning an upgrade path for an Internet-connected Sun environment, this book spends a lot of time documenting the factors that can create performance bottlenecks in Solaris servers and explaining how to measure peak loads. Hourly, monthly, seasonal, and special load increases are modeled mathematically, and the authors offer ideas for planning hardware capacity to match demand. They're explicit in their recommendations, calling for specific processor speeds, disk capacities, bandwidth provisions, and other specifications. Throughout, the emphasis is on keeping server capacity just slightly ahead of user requirements, so as to minimize expense. --David Wall

Topics covered: Models for predicting Internet users' demands upon Sun servers providing database lookups, transaction processing, and other services. Strategies for documenting user demand, tools for making observations, mathematical models for making predictions, and processes for planning capacity are all detailed.