基于订阅内容相似性分组拓扑构建算法

Construction algorithm based on similarity of subscription

现有发布/订阅系统分组普遍采用随机平均分组,没考虑代理的订阅内容会造成事件在所有分组中转发,增加网络流量。为此,提出一种基于订阅内容相似性分组拓扑构建算法。将每个代理的本地订阅聚集为一个订阅表达式,利用订阅间的相似性,把相似订阅所在代理添加到同一个组中,形成分组拓扑结构。实验结果表明,在不增加订阅信息转发流量和事件转发时延的情况下,可以明显降低事件在代理覆盖网之间转发流量,从而降低网络带宽占用率和代理平均负载。

Existing topology grouping algorithms of publish/subscribe systems generally adopt random and average way, have no consideration of local subscriptions of brokers, which result in forwarding the event in all groups, increasing network traffic. Therefore, an algorithm of topology grouping based on similarity of subscriptions is proposed. Firstly, each broker＇s subscription is aggregated into one subscription expression, then similar content of subscription is computed, and similar subscriptions are grouped lastly. Experimental results show that event {orwarding tra{fic, occupancy rate and average load of overlay networks are reduced significantly compared with the existing approaches.