Workload Characterization of a Large Systems Conference Web Server

This paper presents a workload characterization study of the WWW2007 conference Web site. We use both server-sideband client-side measurement to collect data for our analysis. The collected datasets contain approximately 10 million requests from about 130,000 unique visitors generating 215 GB of traffic volume during a 1-year period. Our analysis focuses on usage behaviour, client errors, client network properties,bandwidth, and robot activity of the site.In addition to the workload characterization study itself,our work also offers a comparative discussion of different approaches to Web data analysis, and some insights for future organizers of large systems conferences with a significant Web presence.