Enhancing Java server availability with JAS

Reviewer: Sajjad A. Khan

The JAS software system for monitoring Java programs with high availability and performance requirements is introduced in this paper. JAS is an external lightweight supervisor for Java applications that operates during runtime without any byte code modifications, and without affecting the functionality of an application. The authors have substantiated their claims by using experimental results with a Web proxy, simulating reliability and performance problems under heavy load scenarios. The quantified performance improvement was 33 percent, and request time decreased by 22 percent. Due to the lack of an existing Java Virtual Machine (JVM) with a built-in application supervisor, the authors faced some challenges when designing this system. In section 2, a good comparison of JAS features is provided, which helps put their work into perspective with other implementations. Although quantitative comparison was not possible with benchmark or real-time systems, the authors describe in detail the main features, limitations, and events that can be detected by the current JAS implementation. The JAS seems suitable for detecting problems like frequent running of garbage collector, and when threads exceed their threshold. Configuration of the JAS for a target application is also described in detail; specifically, thread policy specifications, system specifications, and front-end specifications. Overall, this paper contains a great deal of detail on the performance effectiveness of JAS in a typical experimental situation. However, the results obtained in this environment need to be extrapolated with caution to other real life environments, where JVM and hardware speed, the number of thread classes specified in the JAS configuration, and the total number of JAS related events generated by the target application may vary. Overall, this is a great effort to describe a system written for Java servers that has the potential to enhance performance and reliability. Online Computing Reviews Service