The workload, we use to analyze load balancing policies in clustered Web servers, consist of the server logs of the 1998 World Soccer Cup Web site7.1. The World Cup site server was composed of 30 low-latency platforms distributed across four geographically distributed locations. Client requests were dispatched to a location via a Cisco Distributed Director, and each location was responsible for load balancing of incoming requests among its available servers. The traces provide information about each request received by the site during 92 days, from April 26, 1998 to July 26, 1998. Trace data were collected for each day during the total period of time that the Web server was operational.
The traces provide information about each request received by each server. For each request the following information is recorded: the IP address of the client issuing the request, the date and time of the request, the URL requested, the HTTP response status code, and the content length (in bytes) of the transferred document. Trace data were collected for each day during the total period of time that the Web server was operational. Since the focus of this work is on load balancing, irrespective of possible caching policies at the server, we only extracted the content length of the transfered document from each trace record assuming that the service time of each request is a linear function of the size of the requested document. Analysis of the unique file size distribution across all 92 days indicated that the service process in the Web server is heavy-tailed. This trend persists also if the empirical distribution of the sizes of requests files is analyzed on a day by day basis. For a detailed analysis of the World Cup workload see [3].