cs 15(4): e5

Research Article

PETFEN: A Performance Evaluation Tool for Flow-Level Network Modeling of Ethernet Networks

Download982 downloads
  • @ARTICLE{10.4108/icst.valuetools.2014.258166,
        author={Fabien Geyer and Stefan Schneele and Georg Carle},
        title={PETFEN: A Performance Evaluation Tool for Flow-Level Network Modeling of Ethernet Networks},
        journal={EAI Endorsed Transactions on Cloud Systems},
        volume={1},
        number={4},
        publisher={EAI},
        journal_a={CS},
        year={2015},
        month={2},
        keywords={flow-level network modeling, network traffic modeling, performance evaluation},
        doi={10.4108/icst.valuetools.2014.258166}
    }
    
  • Fabien Geyer
    Stefan Schneele
    Georg Carle
    Year: 2015
    PETFEN: A Performance Evaluation Tool for Flow-Level Network Modeling of Ethernet Networks
    CS
    EAI
    DOI: 10.4108/icst.valuetools.2014.258166
Fabien Geyer1,*, Stefan Schneele2, Georg Carle3
  • 1: Airbus Group Innovations and Technische Universität München
  • 2: Airbus Group Innovations
  • 3: Technische Universität München
*Contact email: fabien.geyer@airbus.com

Abstract

We present in this paper PETFEN, a Performance Evaluation Tool for Flow-level network modeling of Ethernet Networks. Flow-level network models are a useful tool to dimension and predict various performances of networks with TCP and UDP flows, providing information such as mean flow bandwidths, link utilizations or queue sizes. While the literature on flow-level network models is extensive, there is still a lack of tools for numerical evaluations on user provided topologies. In this paper, we describe the three components of PETFEN: (i) an effective domain specific language used for algorithmically describing topologies, (ii) a mathematical toolbox for the numerical evaluation of flow-level network models on the provided topologies, (iii) modules for the evaluation of the topologies with external tools. Via various numerical evaluations, we compare the results of PETFEN with results of SimGrid, another tool based on flow-level network models, as well as results of the discrete event simulator OMNeT++.