Memory leak detection in Java: Taxonomy and classification of approaches

Abstract

Memory leaks are usually not associated with runtime environments with automatic garbage collection; however, memory leaks do happen in such environments and present a challenge to detect and find a root cause. Currently in the industry manual heap dump analysis is the most popular way of finding memory leaks, regardless of the number of automated methods proposed by scientists over the years. However, heap dump analysis alone cannot answer all questions needed to fix the leak effectively. The current paper reviews memory leak detection approaches proposed over the years and classifies them from the point of view of assessed metrics, performance overhead and intrusiveness. In addition, we classify the methods into online, offline and hybrid groups based on their features. © 2014 Elsevier Inc.