Adaptive Instrumentation of Java Applications for Experiment-Based Performance Analysis

Schulz, Henning, Flaig, Albert, Wert, Alexander and van Hoorn, Andre (2014) Adaptive Instrumentation of Java Applications for Experiment-Based Performance Analysis [Talk] In: Symposium on Software Performance 2014: Joint Descartes/Kieker/Palladio Days, November 26-28, 2014, Stuttgart, Germany.

[img]
Preview
Slideshow
SOSP-Vortrag-combined20141126.pdf - Presentation

Download (1635Kb) | Preview

Abstract

Running load tests, instrumentation of selected application parts is a common practice in order to measure performance metrics. Instrumentation means to extend the target application by measurement probes while executing measurements. For instance, in Java bytecode instrumentation, additional commands are inserted into the bytecode of the target application. Since data generation is time-consuming, it may affect the target application and thus the measurement. Countering this problem, stepwise approaches execute several measurements while using only few measurement probes per measurement. Utilizing existing approaches, the target application has to be restarted in order to change the instrumentation. The resulting measurement overhead can cause the execution of stepwise measurements to be impracticable or bound to high manual effort. In this presentation, we introduce the Adaptable Instrumentation and Monitoring (AIM) framework enabling stepwise measurements without system restarts. Furthermore, we show the advantages of selective instrumentation with AIM over excessive instrumentations. For instance, we introduce an approach to highly precise and fully automated performance-model calibration. Thereby, the relative error is smaller than 4%, whereas excessive instrumentations introduce an error of up to 50%. Last not least, we present the embedding of AIM in the Kieker framework, merging the adaptability of AIM with the comprehensive monitoring and analysis capabilities of Kieker.

Document Type: Conference or Workshop Item (Talk)
Keywords: Dynamic instrumentation, Kieker, adaptive monitoring, AIM
Research affiliation: Kiel University > Faculty of Engineering > Department of Computer Science > Software Engineering
Projects: Kieker
Date Deposited: 13 Dec 2014 20:21
Last Modified: 17 Dec 2014 23:13
URI: http://eprints.uni-kiel.de/id/eprint/26488

Actions (login required)

View Item View Item

Document Downloads

More statistics for this item...