Quality, Reliability, Security and Robustness in Heterogeneous Systems. 14th EAI International Conference, Qshine 2018, Ho Chi Minh City, Vietnam, December 3–4, 2018, Proceedings

Research Article

Exploring YouTube’s CDN Heterogeneity

  • @INPROCEEDINGS{10.1007/978-3-030-14413-5_13,
        author={Anh-Tuan Nguyen and Olivier Fourmaux and Christophe Deleuze},
        title={Exploring YouTube’s CDN Heterogeneity},
        proceedings={Quality, Reliability, Security and Robustness in Heterogeneous Systems. 14th EAI International Conference, Qshine 2018, Ho Chi Minh City, Vietnam, December 3--4, 2018, Proceedings},
        proceedings_a={QSHINE},
        year={2019},
        month={3},
        keywords={CDN Measurement YouTube PlanetLab EdgeNet},
        doi={10.1007/978-3-030-14413-5_13}
    }
    
  • Anh-Tuan Nguyen
    Olivier Fourmaux
    Christophe Deleuze
    Year: 2019
    Exploring YouTube’s CDN Heterogeneity
    QSHINE
    Springer
    DOI: 10.1007/978-3-030-14413-5_13
Anh-Tuan Nguyen1,*, Olivier Fourmaux1,*, Christophe Deleuze2,*
  • 1: Sorbonne Université, CNRS, LIP6
  • 2: Univ. Grenoble Alpes, Grenoble INP (Institute of Engineering Univ. Grenoble Alpes), LCIS
*Contact email: anh-tuan.nguyen@lip6.fr, olivier.fourmaux@lip6.fr, christophe.deleuze@lcis.grenoble-inp.fr

Abstract

In this paper, we set up measurements and make performance and geographic analysis of YouTube CDN video platform. We use large distributed testbeds, like PlatnetLab and EdgeNet, to grasp the heterogeneity of client situations. Those facilities can work as real clients without any simulation. From these infrastructures, we generate numerous requests to YouTube video servers. Using a reduced initial set of nodes in different geographic location, we continuously measure information related to YouTube homepage websites and video servers, and calculate the latency from clients to cache servers. We also look at the geographical location of YouTube servers. This enables a better understanding of cache mapping strategy and draws the map of the system. Our first result focus on distance between users and data centers before studying dynamic aspect of the system. The information we collect can be of interest to e.g. ISP network operators who need to improve their network architecture to minimize costs and optimize quality for the user.