Marwede, Nina Sophie (2008) Automatic Failure Diagnosis based on Timing Behavior Anomaly Correlation in Distributed Java Web Applications. (Diploma thesis), University of Oldenburg, Oldenburg, Germany, 119 pp.

Preview

Text Marwede2008AutomaticFailureDiagnosisBasedOnTimingBehaviorAnomalyCorrelationInDistributedJavaWebApplications.pdf - Published Version Download (2MB) | Preview

Abstract

One approach to handling software failures, especially in large distributed systems, is the monitoring of components for quick reaction and recovery to reduce downtimes and maintenance costs. In previous work, our group has developed tools to monitor and evaluate the timing behavior of Java software systems to detect anomalies, which may be indicators of erroneous behavior.

This diploma thesis enhances the approach to isolate the root cause for failures in distributed systems. Existing technologies are connected and extended to aggregate anomalies to a failure diagnosis. An anomaly correlator is developed that combines timing behavior anomalies using derived dependency graphs. It is then applied to a sample application under the influence of workload generation and fault injection. The results are visualized as colored graphs that can be exported to various image formats.

Actions (login required)

View Item