Research Talk: Heap Abstraction for Visualization

Abstract

Understanding the architecture and structure of data is crucial to software development and debugging. Existing heap visualization tools suffer from a number of significant limitations. A number of existing tools focus on detecting memory problems, while the others do not scale well when visualizing large heaps.

The Heapviz tool provides an interactive framework for analyzing and visualizing heap snapshots from running Java programs. Heapviz represents heap snapshots as directed graphs and uses a graph summarization algorithm to reduce the size and complexity of the graph. The existing graph summarization algorithm fails to achieve a satisfactory level of abstraction on certain real-world benchmarks such as SPEC JBB.

In this talk, I will present preliminary research on a general abstraction framework that will allow multiple levels of heap abstraction. We expect the proposed framework to improve the compression and abstraction that the current Heapviz summarization algorithm achieves.