拓扑感知和局部负载均衡的应用层组播方案

Topology-aware and local load-balancing application layer for multicast schemes

应用层组播是一种适合大规模流媒体分发的方案。为了解决异质网络条件下Peer节点的负载均衡问题和降低覆盖层传输延迟,提出一个二层构架方案,覆盖网络根据网络距离进行聚类实现拓扑感知来降低覆盖层网络延迟;在聚类网络的内部,通过基于预期适应度的有偏游走算法构建局部负载均衡的覆盖网络。"中国教育与研究网"(Cernet)主干网上的实验验证了节点聚类方法的有效性。仿真结果表明该方案能够以O(lgN)的游走步长得到比现有系统更小的相对负载方差。

The application layer multicast （ALM） scheme is suitable for large live media file distribution. A two-level hierarchal ALM scheme was developed to improve peer load-balancing in heterogeneous networks and decrease the overlay transmission latency. Locality-awareness was used to reduce the overlay delay through peers clustering according to the peer network distances in the overlay networks. Peers in one cluster were self-organized into a local load-balancing overlay based on an expected fitness based preferential random walk. Experiments on CERNET （China education and research network） verify the effectiveness of the peer clustering scheme. Simulations show that the system provides better load-balancing performance than existing systems with a O（lgN） step length random walk.