Home // ICSEA 2013, The Eighth International Conference on Software Engineering Advances // View article

Low-Overhead Profiling based on Stationary and Ergodic Assumptions

Authors:
Stoyan Garbatov
João Cachopo

Keywords: profiling; real-time monitoring; feedback-directed optimizations; performance; ergodic; stationary

Abstract:
In the context of feedback-directed optimization solutions, the component responsible for collecting application behaviour data cannot afford to introduce any performance overheads, otherwise it undermines any optimization that is to be carried out. This work presents a new online solution for profiling object-oriented applications. The solution collects detailed information about accesses over domain instances and their fields, while introducing approximately zero overheads. This is accomplished by making assumptions about the stationary and ergodic properties of applications' run-time behaviour. The work has been validated with the TPC-W benchmark.

Pages: 380 to 386

Copyright: Copyright (c) IARIA, 2013

Publication date: October 27, 2013

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-61208-304-9

Location: Venice, Italy

Dates: from October 27, 2013 to October 31, 2013