基于超节点网络的多源并行数据分发研究

在P2P系统中,同一个数据往往拥有多个副本,这些副本分布在网络的不同节点上,为并行分发提供了可能.然而在传统的数据分发中,只有单一的源节点参与分发,这种模式不但浪费了系统中存在的多个副本资源,而且容易使单一的源节点成为系统的瓶颈.为了更有效地利用系统中存在的多个副本资源,提高分发的效率和系统的性能,提出了一个基于超节点网络的多源并行数据分发算法PPMSD,通过一个基于超节点网络的结构化拓扑来组织系统中的数据和索引,使得数据的查询和定位效率控制在O(log N)以内;根据分发速率动态地为每个数据源分配相应比例的分发数据量,并行地向目标节点传输数据以最小化分发时间;最后,实现了原型系统PPThor以验证算法的有效性.测试结果表明分发的效率得到了极大的提高.

Research on Multi-Source Parallel Data Dissemination Based on Super-Peer Network

In P2P systems,one data object often owns several replicas in the network,which gives a chance to parallel data dissemination.However,the traditional dissemination systems which distribute data from one source to the target do not take full advantage of these replicas.Moreover,the single-source schema may result in single point failure and hot spot bottleneck.In order to make the best of multiple replicas and improve the dissemination efficiency,a new multi-source parallel data dissemination algorithm named as PPMSD is proposed.It uses a super-peer based network to organize the data and index.SO that the query can be processed with O(log N)efficiency.The algorithm allocates the ration of data to each source in proportion to its dissemination speed and makes multiple replicas transfer data to the target in parallel so as to reduce the dissemination time.At last,a prototype system is implemented to evaluate the algorithm and the result shows that the efficiency of the system is greatly improved.