CDACAN: a scalable structured P2P network based on continuous discrete approach and CAN

  • Authors:
  • Lingwei Li;Qunwei Xue;Deke Guo

  • Affiliations:
  • School of Information System and Management, National University of Defense Technology, Changsha, Hu Nan, P.R. China;China Institute of Geo-Environment Monitoring, Beijing, China;School of Information System and Management, National University of Defense Technology, Changsha, Hu Nan, P.R. China

  • Venue:
  • HPCC'07 Proceedings of the Third international conference on High Performance Computing and Communications
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

CAN is a famous structured peer-to-peer network based on d-dimensional torus topology with constant degree and logarithmical diameter, but suffers from poor scalability when N≫2d, N is the number of peers. To address this issue, we proposes a novel scalable structured peer-to-peer overlay network, CDACAN that embeds the one-dimensional discrete distance halving graph into each dimension of CAN. The out-degree and average routing path length of CDACAN are O(d) and O(log (N)), respectively, and are better than that of CAN. On the other hand, we analyze the optimal value of dimensions and the smooth division method of d-dimensional Cartesian coordinate space when handling the dynamic operations of peers. The smooth division of multidimensional space can improve the routing performance, and also is helpful to keep load balance among peers. Those properties and protocols are carefully evaluated by formal proofs or simulations. Furthermore, we present a layered improving scheme to decrease the out-degree of each peer in the future work. The expected topology will keep 8 out-degree and O(log2(N)+d) routing path length.