Abstract:

Java's speed is now on par with C/C++ thanks to recent improvements in Java class libraries and Just-in-Time (JIT) compilation methods. An appealing objective is to make the Java Virtual Machine (JVM) "cluster-aware" in order to fully utilize Java's multithreading feature on clusters. This would allow a collection of JVMs operating on dispersed nodes to function as a single, more potent JVM, enabling true parallel execution of a multithreaded Java application. However, there hasn't been much research done on using DJVMs to cluster real-world web applications with workloads from commercial servers. In this article, a novel generic clustering method based on DJVMs that encourages global object sharing and user transparency for web application servers is presented. The optimization methods recently used for the Just-In-Time compilers included in the IBM® Developer Kit for JavaTM and the J9 Java virtual machine standard are described in this document. It mostly concentrates on the improvements that enhanced the performance of the middleware and servers. Virtual machines (VMs) and just-in-time (JIT) compilers face a number of performance issues when it comes to large server and middleware applications built in the Java programming language; we need to handle both steady-state performance and startup time. We outline 12 enhancements that have been made to IBM products in this article because they enhance the scalability and performance of these kinds of applications. For instance, these optimizations lower the cost of object creation, synchronization, and other frequently used Java class library calls. We also outline methods, including recompilation strategies, to deal with server start-up time. According to the experimental findings, the modifications we go over in this article may increase the performance of programs like SPECjbb2000 and SPECjAppServer2002 by up to 10% to 15%.

Description:

.