7th International Conference on Performance Evaluation Methodologies and Tools

Research Article

Practical Performance Evaluation of Ethernet Networks with Flow-Level Network Modeling

  • @INPROCEEDINGS{10.4108/icst.valuetools.2013.254367,
        author={Fabien Geyer and Stefan Schneele and Georg Carle},
        title={Practical Performance Evaluation of Ethernet Networks with Flow-Level Network Modeling},
        proceedings={7th International Conference on Performance Evaluation Methodologies and Tools},
        publisher={ICST},
        proceedings_a={VALUETOOLS},
        year={2014},
        month={1},
        keywords={flow-level network modeling network traffic modeling performance evaluation},
        doi={10.4108/icst.valuetools.2013.254367}
    }
    
  • Fabien Geyer
    Stefan Schneele
    Georg Carle
    Year: 2014
    Practical Performance Evaluation of Ethernet Networks with Flow-Level Network Modeling
    VALUETOOLS
    ACM
    DOI: 10.4108/icst.valuetools.2013.254367
Fabien Geyer1,*, Stefan Schneele1, Georg Carle2
  • 1: EADS Innovation Works
  • 2: Technische Universitaet Muenchen
*Contact email: fabien.geyer@eads.net

Abstract

Network models for evaluating the behavior of networks are important tools in traffic engineering for dimensioning networks and provisioning bandwidth for different applications. We present in this paper a flow-level network model for the performance evaluation of IP networks with support of long-lived TCP and UDP flows. While flow-level network models for TCP and UDP flows have already been investigated, a vast majority of previous studies often do not take into account the importance of cross-traffic. This paper presents topologies where cross-traffic has a major impact on the performance of TCP flows and shows how previous models are not accurate enough. We consider in our study Ethernet LANs with low latencies and show how to apply our framework to networks with Ethernet switches using priority based scheduling, fair-queuing scheduling, or hierarchical scheduling based on the former algorithms. We assess the accuracy of our approach by comparing the results of our model with results of the discrete event simulator OMNeT++.