订阅/发布模式下多节点协同的数据分发方法

Data distribution method based on multi-node cooperation in subscribe /publish model

为了提高分布对称体系结构的订阅/发布系统对大数据量内容的分发效率,提出了一种基于订阅节点协同的数据分发方法.首先,利用MD5算法将订阅节点映射到32 bit逻辑地址空间中;然后依据订阅者与发布者间的逻辑距离所处区间,将订阅节点集合划分成独立不相交的桶,为主题数据的转发规划出合理的路径,且该数据分发路径能适应系统的动态变化.基于逻辑距离的桶分割方法确保了数据分发的单向收敛性,限制了分发路径的深度.真实环境中的实验结果表明,基于订阅节点协同的分发方式通过利用订阅节点的资源,降低了对GB级数据量内容进行分发的分发时延,减轻了发布节点的负载,与传统的点对点分发方式相比,数据分发效率得到了显著提高.

To improve the distribution efficiency of bulk content in the subscribe/publish system with symmetrical distributed architecture,a new distribution method based on multi-node cooperation is proposed.First,the subscribers are mapped to a 32 bit logical address space by the message-digest al-gorithm 5 （MD5 algorithm）.Then,the sets of the subscribers are divided into independent disjoint buckets according to the intervals of the logical distances between the publisher and the subscribers. A proper route for data forwarding,which can adapt to the dynamic changes in the system,is gener-ated.The unidirectional convergence of data distribution and the limited depth of the route are guar-anteed by the bucket partitioning method based on the logical distance.The prototype experimental evaluation results show that the proposed method reduces subscribers＆#39; delay and publisher＆#39;s load for the distribution of bulk content in GB size by exploiting subscribers＆#39;resources.This method outper-forms the traditional point-to-point distribution strategy in distribution efficiency.