Testbeds and Research Infrastructures. Development of Networks and Communities. 6th International ICST Conference, TridentCom 2010, Berlin, Germany, May 18-20, 2010, Revised Selected Papers

Research Article

Experimental Evaluation of OpenVZ from a Testbed Deployment Perspective

Download
559 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-17851-1_7,
        author={Gautam Bhanage and Ivan Seskar and Yanyong Zhang and Dipankar Raychaudhuri and Shweta Jain},
        title={Experimental Evaluation of OpenVZ from a Testbed Deployment Perspective},
        proceedings={Testbeds and Research Infrastructures. Development of Networks and Communities. 6th International ICST Conference, TridentCom 2010, Berlin, Germany, May 18-20, 2010, Revised Selected Papers},
        proceedings_a={TRIDENTCOM},
        year={2012},
        month={10},
        keywords={},
        doi={10.1007/978-3-642-17851-1_7}
    }
    
  • Gautam Bhanage
    Ivan Seskar
    Yanyong Zhang
    Dipankar Raychaudhuri
    Shweta Jain
    Year: 2012
    Experimental Evaluation of OpenVZ from a Testbed Deployment Perspective
    TRIDENTCOM
    Springer
    DOI: 10.1007/978-3-642-17851-1_7
Gautam Bhanage1,*, Ivan Seskar1,*, Yanyong Zhang1,*, Dipankar Raychaudhuri1,*, Shweta Jain1,*
  • 1: Rutgers University
*Contact email: gautamb@winlab.rutgers.edu, seskar@winlab.rutgers.edu, yyzhang@winlab.rutgers.edu, ray@winlab.rutgers.edu, sjain@winlab.rutgers.edu

Abstract

A scalable approach to building large scale experimentation testbeds involves multiplexing the system resources for better utilization. Virtualization provides a convenient means of sharing testbed resources among experimenters. The degree of programmability and isolation achieved with such a setup is largely dependent on the type of technology used for virtualization. We consider OpenVZ and User Mode Linux (UML) for virtualization of the ORBIT wireless testbed and evaluate their relative merit. Our results show that OpenVZ, an operating system level virtualization mechanism significantly outperforms UML in terms of system overheads and performance isolation. We discuss both qualitative and quantitative performance features which could serve as guidelines for selection of a virtualization scheme for similar testbeds.