Using a Queuing Model to Analyze the Performance of Web Servers

Despite the increasing number of Web (i.e., HTTP) servers in use each day, little is definitively known about their performance characteristics. In this paper, we present a simple, high-level, open queuing network model from which we derive several general performance results for web servers on the Internet. Also, we analyze multiple-server systems. A theoretical upper bound on the serving capacity of web servers is defined. As Web servers approach this boundary, response time increases suddenly towards infinity, which disables the server. Limiting the server's simultaneous connections prevents this problem. The effects of file size, server speed, and network bandwidth on response time are also investigated. In addition, the relative merits of several methods of improving server performance are evaluated.