Complex Sciences. First International Conference, Complex 2009, Shanghai, China, February 23-25, 2009, Revised Papers, Part 2

Research Article

Cache Allocation in CDN: An Evolutionary Game Generalized Particle Model

Download
429 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-02469-6_4,
        author={Xiang Feng and Francis Lau and Daqi Gao},
        title={Cache Allocation in CDN: An Evolutionary Game Generalized Particle Model},
        proceedings={Complex Sciences. First International Conference, Complex 2009, Shanghai, China, February 23-25, 2009, Revised Papers, Part 2},
        proceedings_a={COMPLEX PART 2},
        year={2012},
        month={5},
        keywords={Content Delivery Networks (CDN) cache resource allocation evolutionary game generalized particle model (G-GPM) placement algorithm distributed and parallel algorithm},
        doi={10.1007/978-3-642-02469-6_4}
    }
    
  • Xiang Feng
    Francis Lau
    Daqi Gao
    Year: 2012
    Cache Allocation in CDN: An Evolutionary Game Generalized Particle Model
    COMPLEX PART 2
    Springer
    DOI: 10.1007/978-3-642-02469-6_4
Xiang Feng1,*, Francis Lau2,*, Daqi Gao1,*
  • 1: East China University of Science and Technology
  • 2: The University of Hong Kong
*Contact email: xfeng@ecust.edu.cn, fcmlau@cs.hku.hk, gaodaqi@ecust.edu.cn

Abstract

Content distribution networks (CDNs) increasingly have been used to reduce the response times experienced by Internet users through placing surrogates close to the clients. This paper presents an object replacement approach based on an evolutionary game generalized particle model (G-GPM). We first propose a problem model for CDNs. The CDN model is then fit into a gravitational field. The origin servers and surrogates are regarded as two kinds of particles which are located in two force-fields. The cache allocation problem is thus transformed into the kinematics and dynamics of the particles in the annular and the round force-fields. The G-GPM approach is unique in four aspects: 1) direct viewing of individual and overall optimization; 2) parallel computing (lower time complexity); 3) multi-objective solution; and 4) being able to deal with some social interactions behaviors.