Further investigations into evaluating Web site reliability

This paper explores the idea of evaluating Web site reliability based on the data from server logs. This approach requires the extraction of workload measures and the error types to be extracted from the server logs. This study follows on another study, and hence, can be regarded as a partial replication of the original study. While the approach in the original study is feasible, we have to question the effectiveness of just using the 404 error type and hit count to estimate the reliability of Web sites. Various error types are examined and their implications when used for reliability analysis are discussed. Potential issues with the workloads defined in the original study are also investigated and discussed. After a thorough investigation, we believe that reliability analysis for Web sites can and should be extended to include other error codes and workloads as they can provide a much better reliability estimate of today´s highly dynamic Web sites.