Model-based performance analysis using block coverage measurements

Swapna S. Gokhale

doi:10.1016/j.jss.2008.06.051

Return

Model-based performance analysis using block coverage measurements

2009-01-01

PRE

Swapna S. Gokhale *

DOI：10.1016/j.jss.2008.06.051

Original

Original request for help

Save

Abstract

En 中文

The primary advantage of model-based performance analysis is its ability to facilitate sensitivity and predictive analysis, in addition to providing an estimate of the application performance. To conduct model-based analysis, it is necessary to build a performance model of an application which represents the application structure in terms of the interactions among its components, using an appropriate modeling paradigm. While several research efforts have been devoted to the development of the theoretical aspects of model-based analysis, its practical applicability has been limited despite the advantages it offers. This limited practical applicability is due to the lack of techniques available to estimate the parameters of the performance model of the application. Since the model parameters cannot be estimated in a realistic manner, the results obtained from model-based analysis may not be accurate. In this paper, we present an empirical approach in which profile data in the form of block coverage measurements is used to parameterize the performance model of an application. We illustrate the approach using a network routing simulator called Maryland routing simulator (MaRS). Validation of the performance estimate of MaRS obtained from the performance model parameterized using our approach demonstrates the viability of our approach. We then illustrate how the model could be used for predictive performance analysis using two scenarios. By the virtue of using code coverage measurements to parameterize a performance model, we integrate two mature, yet independent research areas, namely, software testing and model-based performance analysis.. (c) 2008 Elsevier Inc. All rights reserved.

Keywords:

Model-based performance analysis

Profile data

Test coverage

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.