Root-cause analysis of performance anomalies in web-based applications

Abstract

The complexity behind current business-critical applications leads many times to performance problems difficult to anticipate and analyze. In our previous work we described a framework for detection of performance anomalies in web-based and component-based applications. It provides low overhead monitoring, correctly distinguishes performance anomalies from common workload variations and also presents initial information for system or application server changes related with an application performance anomaly. In this paper we present a framework extension devised to offer root-cause failure analysis for a given performance anomaly. The monitoring module enables application profiling and ANOVA analysis is used to verify if a performance anomaly is due to internal changes within the application (e.g., application updates) or to external changes (e.g., remote services changes, system/application server change). The paper includes some experimental results that show the effectiveness of our approach to pinpoint the root-cause for different types of performance anomalies and remarks its potential to avoid a considerable number of service failures.