Automatic failure detection, logging, and recovery for high-availability Java servers

Abstract

Many systems and techniques exist for detecting application failures. However, previously known generic failure detection solutions are only of limited use for Java applications because they do not take into consideration the specifics of the Java language and the Java execution environment. In this article, we present the application-independent Java Application Supervisor (JAS). JAS can automatically detect, log, and resolve a variety of execution problems and failures in Java applications. In most cases, JAS requires neither modifications nor access to the source code of the supervised application. A set of simple user-specified policies guides the failure detection, logging, and recovery process in JAS. A JAS configuration manager automatically generates default policies from the bytecode of an application. The user can modify these default policies as needed. Our experimental studies show that JAS typically incurs little execution time and memory overhead for the target application. We describe an experiment with a Web proxy that exhibits reliability and performance problems under heavy load and demonstrate an increase in the rate of successful requests to the server by almost 33% and a decrease in the average request processing time by approximately 22% when using JAS.