• DocumentCode
    1801467
  • Title

    How can multi-application co-exist dependably and efficiently in CSDN?

  • Author

    Peichang Shi ; Huaimin Wang ; Gang Yin ; Liu, Xuening ; Xiaoqun Yuan ; Dianxi Shi

  • Author_Institution
    School of Computer and Science, National University of Defense Technology, Hunan, China
  • fYear
    2013
  • fDate
    1-8 Jan. 2013
  • Firstpage
    1
  • Lastpage
    6
  • Abstract
    Providing optimal Quality of Service (QoS) for Cloud Services Delivery Networks (CSDN) is a challenging task due to the mutual-interference phenomenon among multiple applications. Existing approaches only provide best-effort services, which aim to maximize the number of served users and sacrifice service dependability. To enhance the service dependability, designs of allocating fixed bandwidth among different applications are proposed, which however, lead to imbalanced bandwidth utilization among these applications. It implies that some applications have surplus bandwidth, while some others suffer low service quality due to bandwidth deficit. In this paper, we propose a method named Multi-Application Co-Exist (MACE), considering the trade-off between service dependability and bandwidth utilization efficiency. In MACE, the mutual-interference multi-application are classified and deployed isolatedly to some extent. We evaluate our MACE approach with 6-month traces obtained from a productive CSDN. Compared with existing approaches, the average bandwidth utilization of all applications has been increased by about 20% and the average statistic multiplexing ratio of all applications has been decreased from 12% to 5%. Moreover, the number of complaint events affecting the dependability of CSDN services caused by mutual-interference among different applications has been dropped to 0.
  • Keywords
    Bandwidth; Business; Equations; Interference; Measurement; Multiplexing; Servers; Cloud services; delivery network; dependable; efficient; multi-application;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Conference Anthology, IEEE
  • Conference_Location
    China
  • Type

    conf

  • DOI
    10.1109/ANTHOLOGY.2013.6784796
  • Filename
    6784796