10th EAI International Conference on Simulation Tools and Techniques

Research Article

On Improving Parallel Real-Time Network Simulation for Hybrid Experimentation of Software Defined Networks

  • @INPROCEEDINGS{10.1145/3173519.3173535,
        author={Mohammad Obaida and Jason Liu},
        title={On Improving Parallel Real-Time Network Simulation for Hybrid Experimentation of Software Defined Networks},
        proceedings={10th EAI International Conference on Simulation Tools and Techniques},
        publisher={ACM},
        proceedings_a={SIMUTOOLS},
        year={2018},
        month={8},
        keywords={real-time network simulation parallel simulation emulation software-defined networking sdn},
        doi={10.1145/3173519.3173535}
    }
    
  • Mohammad Obaida
    Jason Liu
    Year: 2018
    On Improving Parallel Real-Time Network Simulation for Hybrid Experimentation of Software Defined Networks
    SIMUTOOLS
    ACM
    DOI: 10.1145/3173519.3173535
Mohammad Obaida1, Jason Liu1,*
  • 1: Florida International University
*Contact email: liux@cis.fiu.edu

Abstract

Real-time network simulation enables simulation to operate in real time, and in doing so allows experiments with simulated, emulated, and real network components acting in concert to test novel network applications or protocols. Real-time simulation can also run in parallel for large-scale network scenarios, in which case network traffic is represented as simulation events passed as messages to remote simulation instances running on different machines. We note that substantial overhead exists in parallel real-time simulation to support synchronization and communication among distributed instances, which can significantly limit the performance and scalability of the hybrid approach. To overcome these challenges, we propose several techniques for improving the performance of parallel real-time simulation, by eliminating parallel synchronization and reducing communication overhead. Our experiments show that the proposed techniques can indeed improve the overall performance. In a use case, we demonstrate that our hybrid technique can be readily integrated for studies of software-defined networks.