Study of Server Workload in Large Online Newspaper

In this paper we present an analysis of access traces from a large online newspaper offering both web and Video-on- Demand service. We have analyzed the lifetime of news articles, their popularity, server load, session interactivity, the working set of servers, and the correlation of the number of web and streaming accesses within a session. The results are interpreted in the context of designing a content distribution network targeted at News-on-Demand applications. We show that all of the requested web and streaming content could fit in the main memory of a low-end server.