Testbeds and Research Infrastructure. Development of Networks and Communities. 8th International ICST Conference, TridentCom 2012, Thessanoliki, Greece, June 11-13, 2012, Revised Selected Papers

Research Article

DNEmu: Design and Implementation of Distributed Network Emulation for Smooth Experimentation Control

Download
506 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-35576-9_16,
        author={Hajime Tazaki and Hitoshi Asaeda},
        title={DNEmu: Design and Implementation of Distributed Network Emulation for Smooth Experimentation Control},
        proceedings={Testbeds and Research Infrastructure. Development of Networks and Communities. 8th International ICST Conference, TridentCom 2012, Thessanoliki, Greece, June 11-13, 2012, Revised Selected Papers},
        proceedings_a={TRIDENTCOM},
        year={2012},
        month={12},
        keywords={distributed emulation real-time simulation ns-3},
        doi={10.1007/978-3-642-35576-9_16}
    }
    
  • Hajime Tazaki
    Hitoshi Asaeda
    Year: 2012
    DNEmu: Design and Implementation of Distributed Network Emulation for Smooth Experimentation Control
    TRIDENTCOM
    Springer
    DOI: 10.1007/978-3-642-35576-9_16
Hajime Tazaki1, Hitoshi Asaeda2
  • 1: National Institute of Information and Communications Technology (NICT)
  • 2: Keio University

Abstract

Conducting a realistic network experiment involving globally distributed physical nodes under heterogeneous environment introduces a requirement of experimentation control between the real world network and emulated/simulated networks. However, there is a gap between them to deploy network experiments. In this paper, we propose the () to fill the gap for the requirements of a planetary-scale network experiment. addresses the issue of real-time execution with message synchronization through distributed processes, and enables us to evaluate protocols with actual background traffic using a fully controlled distributed environment. Through evaluation with micro-benchmarks, we find that our prototype implementation is similar in terms of packet delivery delay and throughput to the existing non-virtualized environment. We also present a use-case of our proposed architecture for a large distributed virtual machine service in a simple control scenario involving actual background traffic on the global Internet. will contribute to research in protocol evaluation and operation in a huge network experiment without interfering with the existing infrastructure.