Home // ICONS 2013, The Eighth International Conference on Systems // View article

Toward System Level Performance Evaluation of distributed Embedded Systems

Authors:
Jukka Saastamoinen
Subayal Khan
Jyrki Huusko
Juha Korpi
Kari Tiensyrjä
Jari Nurmi

Keywords: performance model, application model, platform model, kahn process networks, Y-Chart, UML, SYstemC

Abstract:
In order to manage the increasing complexity and heterogeneity of the distributed embedded systems, the system level performance evaluation must be performed at an early design phase using abstract models of the platforms and applications. Thus, the system level performance simulation techniques for embedded systems play a key role in the architectural exploration phase. The salient performance evaluation methodologies are developed around some key concepts and employ a variety of modelling styles, models of computation and programming languages for performance modelling of embedded systems. These aspects of a performance evaluation methodology might limit it to certain domain(s) of embedded systems and therefore must be investigated. In order to span different domains of distributed systems, a methodology must provide the models for technologies such as communication protocols and middleware technologies which are employed in different domains of distributed embedded systems. Once achieved, the methodology will be able to provide an estimate of the contribution of these technologies in the non-functional properties of the distributed system. This is important since the goal of the survey is to shortlist the performance modelling methodologies which are more feasible for the performance evaluation of different domains of distributed systems. The abstraction level used to model these protocols must be investigated since the accuracy of the related performance numbers depend on the abstraction level used to model these protocols. Furthermore, the increasing complexity of distributed applications demands that the application design phase shall act as a starting point for application workload modelling to reduce the time and effort involved in performance modelling. After comparing the salient methodologies on the basis of modelling style, tools and languages, targeted domain etc., we shortlisted the contributions which are more feasible for performance evaluation of different domains of distributed systems. Afterwards, we describe the models and tools needed by the shortlisted methodologies in order to span the different domain of distributed systems. The article acts as a reference for researchers and industries involved in developing methods and tools for system level performance simulation.

Pages: 169 to 178

Copyright: Copyright (c) IARIA, 2013

Publication date: January 27, 2013

Published in: conference

ISSN: 2308-4243

ISBN: 978-1-61208-246-2

Location: Seville, Spain

Dates: from January 27, 2013 to February 1, 2013