Comparison of instrumentation techniques for dynamic program analysis on the Java virtual machine

Abstract

Dynamic Program Analysis (DPA) tools commonly rely on bytecode instrumentation which is done by means of low-level bytecode manipulation libraries. While providing a lot of flexibility and expressiveness to developers, using those libraries is usually time consuming, error-prone, and tedious. Approaches that leverage aspect-oriented programming (AOP) serve to reduce the development effort by providing higher abstraction layer for creating instrumentations. However, AOP-based tools have certain known limitations, high performance overhead being one of them. This work is intended to help DPA tool developers to choose the proper instrumentation framework that fully satisfies their needs.