Causal Order Multicast Protocol Using Different Information from Brokers to Subscribers
Chayoung Kim, Jinho Ahn · Advanced science and technology letters · 2014
Recently, in Publish/Subscribe (P/S) systems, there has been increasing emphasis in managing end-to-end message delivery performance and message order-based consistency, which are addressed in distributed collaborative applications for on-the-fly data. A causal message ordering is more useful for these distributed applications in which a large number of processes request collaboratively and interactively in services on P/S systems. Also, in P/S systems of wireless sensor networks (WSNs), data fusion, the process of correlating individual sensor readings originating from various nodes into high-level sensing results, depends on the time of occurrence of fused sensor readings, such as causal ordering. In this paper, we present two versions of causal ordering protocols. In the first protocol, only the information of the predecessors immediately before the message piggybacked on each multicast message is transmitted from brokers to subscribers through gossip-style disseminations based on local views for causal ordering. In the second protocol, every sensor broker disseminates the multicast message piggybacked with the latest time-stamped information that represents the gossip round in which the message is generated to subscribers by using global view of gossiping for causal ordering. The features of these two versions might be highly scalable and suitable for the area of the applications requiring only the minimum causal information of message delivery with flexible consistency.