Empirical Comparison of Techniques for Automated Failure Diagnosis
Failures of Internet services and enterprise systems lead to user dissatisfaction and considerable loss of revenue. Manual diagnosis of these failures can be laborious, slow, and expensive. This problem has received considerable attention recently from researchers in systems, databases, machine learning and knowledge discovery, computer architecture, and other areas. The research has been directed at developing automated, efficient, and accurate techniques for diagnosing failures using system monitoring data.