Statistical framework for ns-3: Terminating simulation and regression analysis
2014 (English)In: World Journal of Modelling and Simulation, ISSN 1746-7233, Vol. 10, no 2, 116-125 p.Article in journal (Refereed) Published
The ns-3 simulator is constantly gaining popularity. It plays a central role in many recent research experiments. The accuracy of the reported simulation results of these experiments is an important concern. Thus, the support in ns-3 for simulation methodologies which guarantee the accuracy of these results is a necessity. Also, the simulation results are affected by numerous scenario parameters. The correlation between the results and the simulation parameters is a significant point of interest in many experiments. In this paper, we present a ns-3 statistical framework. It enables calculation of statistically accurate simulation results by applying the terminating simulation methodology. It features simultaneous execution of simulation scenarios in multi-processor and distributed environments. Also, we integrate support for regression analysis procedures. The proposed framework supports linear and polynomial regression analysis models.We consider simulation results as dependent variables and simulation parameters as independent variables. Regression analysis enables identification of simulation scenario parameters which are significantly correlated to given simulation results. Once a valid regression model is found, it is possible to estimatively predict metric values based on simulation parameter values. This may result in substantially reduced time and effort spent on simulation experimentation.
Place, publisher, year, edition, pages
2014. Vol. 10, no 2, 116-125 p.
IdentifiersURN: urn:nbn:se:kth:diva-161043ScopusID: 2-s2.0-84900323852OAI: oai:DiVA.org:kth-161043DiVA: diva2:794429
QC 201503112015-03-112015-03-062015-03-11Bibliographically approved