Java Performance: The Definitive Guide

This book is a classic, practical guide to Java performance tuning, focused on Java 7 and Java 8, systematically covering every layer that affects the performance of Java applications. It begins with methodology, setting out four basic principles of performance testing: test real applications, understand batch elapsed time versus throughput and response time, use statistical methods to cope with performance variation, and test early and often, and analyzes the biases that microbenchmarks tend to introduce. It then introduces operating-system-level CPU, disk, and network analysis tools, the monitoring tools bundled with the JDK such as jcmd, jconsole, jstat, and jmap, and the use of sampling profilers, profiling profilers, Java Mission Control, and Java Flight Recorder. The core chapters center on the JIT compiler, explaining Client/Server compilation modes, code cache, compilation thresholds, inlining, escape analysis, deoptimization, and the tuning trade-offs of tiered compilation; then, progressing from the basics to the advanced, it thoroughly covers garbage collection, including the principles and parameter tuning of the throughput collector, CMS, and G1, and the settings for heap size, generation spaces, metaspace, promotion, and survivor spaces. The memory section discusses heap memory analysis and object lifecycle management (object reuse, weak references, and soft references), as well as native memory footprint, NIO buffers, native memory tracking, huge pages, and compressed oops. The book closes with coverage of thread pools and ForkJoinPool, the cost of synchronization and false sharing, biased locking and spin locks, Web containers and EJB tuning in Java EE, XML/JSON processing, serialization, JDBC and JPA database performance, and subtle pitfalls in the Java SE API such as buffered I/O, class loading, strings, logging, and collections.