Comparison of garbage collectors in Java programming language

Abstract

Considering the need for continuous and uninterrupted service of modern software applications, it is valuable to analyze how garbage collection (GC) algorithms are handling memory challenges. Widely adopted general-purpose programming languages, like Java, represent an inevitable foundation for many modern application developments. In Java Platform, Standard Edition, and accompanying Java Virtual Machine (JVM), several GCs could be used. In the latest version 9.0.1 of Java SE Development Kit (JDK) default GC was changed to Garbage-First (G1) GC, now becoming widely adopted in addition to previously used Parallel GC and Concurrent Mark & Sweep (CMS) GC. Since GC is a vital part of the JVM, changes and upgrades to its implementation, which reflect upon performance results, are properties worth exploring. Using benchmarks to create non-trivial memory pressures, and with extensive data monitoring, this paper analyzes insights gathered about critical performance factors across several GC algorithms. With the evaluation of benchmark elements, such as object allocations from young area to old area and the duration of the collection time, it was possible to compare GC behavior and assess the overall memory management. This paper presents our preliminary research performed in an academic environment on several benchmark cases and our conclusion about it.