Chapter 22. Project: Server Performance

Large-scale client-server architecture is ubiquitous on the Internet. Web sites may service thousands of simultaneous clients, with individual servers processing hundreds of clients . Parallelism can be achieved by multiple processes, by multiple threads within a process, by asynchronous I/O and events within a single process thread or by combinations of these approaches. This chapter explores the interaction of threading, forking, network communication and disk I/O on the performance of servers.

Objectives

• Learn to measure times and control timing errors

• Experiment with server disk I/O performance

• Explore tradeoffs between threads and processes

• Use the POSIX thread libraries

• Understand different threaded-server architectures