4th International ICST Conference on Simulation Tools and Techniques

Research Article

System for Transport Protocol Evaluation with Automatic Calculation of Statistical Accuracy and Distributed and Parallel Execution

Download581 downloads
  • @INPROCEEDINGS{10.4108/icst.simutools.2011.245532,
        author={Aleksandar Milenkoski and Biljana Stojcevska and Oliver Popov},
        title={System for Transport Protocol Evaluation with Automatic Calculation of Statistical Accuracy and Distributed and Parallel Execution},
        proceedings={4th International ICST Conference on Simulation Tools and Techniques},
        publisher={ICST},
        proceedings_a={SIMUTOOLS},
        year={2012},
        month={4},
        keywords={TCP evaluation distributed execution statistical accuracy parallel execution},
        doi={10.4108/icst.simutools.2011.245532}
    }
    
  • Aleksandar Milenkoski
    Biljana Stojcevska
    Oliver Popov
    Year: 2012
    System for Transport Protocol Evaluation with Automatic Calculation of Statistical Accuracy and Distributed and Parallel Execution
    SIMUTOOLS
    ICST
    DOI: 10.4108/icst.simutools.2011.245532
Aleksandar Milenkoski1,*, Biljana Stojcevska1, Oliver Popov2
  • 1: School of Computer Science and Information Technology, University American College Skopje
  • 2: Department of Computer and Systems Sciences (DSV), Stockholm University
*Contact email: milenkoski@uacs.edu.mk

Abstract

The paper deals with the architecture and the performance of a system for gathering and processing simulation data where use case are communications transport protocols. The system is based on the ns-2 network simulator and the tcpeval tool for the evaluation of TCP and TCP related protocols. The work promotes the concept of a controlled simulation replication as central to achieve statistical accuracy of the simulation results. The later assures that in each case of the specific set of simulation runs, their number does not exceed the minimal one required by the desired and predefined accuracy. Moreover, the use of inter-process and inter-thread communication provided by Open MPI and OpenMP makes the execution of the system possible over a multiprocessor distributed architecture that eventually reduced the time needed to achieve the preferred precision. While the later work is in an early stage, the initial results of the benchmark tests indicate significant gains in time based on the metrics native to tcpeval.