Research On Reliable Data Dissemination Technology Under Emergency Environment

With the rapid growing requirement of information sharing, data dissemination inemergency environment is widely applied in scenes like emergency management,disaster relief, intelligent traffic management and network-centric warfare. The commonrequirement of these applications is to delivery messages reliably and timely from thesource node to the nodes that are interest in these messages in heterogeneous anddynamic networks. In emergency environment, large amounts of data are distributed ina short time, which results in network congestion and message loss. In addition, thehighly dynamic network brings a large number of node failures instantaneously, whichposes a higher requirement to the reliability of data dissemination in emergencyenvironment. To achieve reliable data dissemination in emergency environment, Thisthesis focuses on three aspects: routing on data dissemination, node failure detection, aswell as message loss detection and recovery.The routing in data dissemination indicates that selecting communication paths toforward messages from the source node to the target nodes, which is the key techniqueto ensure timely and reliable delivery. To further balance the routing reliability androuting speed for the existing routing methods in dealing with data loss caused by thelink congestion, a reliable hierarchical and hybrid overlay is built, based on which areliable routing method based on the dynamic multicast tree called RRDMT is proposed.RRDMT adopts prefix matching strategy based on a dynamic tree according to the nodenumbers, to reduce the routing latency and to ensure the load-balance of all the nodes.RRDMT employs a rapid forwarding mechanism based on prefix matching to quicklyreconstruct the dissemination paths of the dynamic multicast tree, so as to improve thereliability of the routing. Experimental results confirm that compared with the existingmethods, RRDMT can ensure real-time and reliable routing even in the case that a largepart of links are disabled.The node failure detection in data dissemination can avoid to forward messages tofailure nodes, improve dissemination reliability and real-time performance, and reducecommunication redundancy. To solve the problem of existing node failure detectionmethods in failing detecting all the failed node fast when large-scale nodes fail, a nodefailure detection method based on neighbor exchanging and failure area predictioncalled NEFAP is proposed. Firstly, to achieve the goal of fast failure node detection,NEFAP gradually removes the oldest nodes based on the neighbor exchanging.Secondly, a neighbor exchanging method attached with the records of failure nodes isproposed, in which nodes exchange the records of failure nodes when they exchangeneighbor information to accelerate the spread speed of failure node messages in thenetwork to improve the detection speed. Thirdly, a failure area prediction strategy isproposed, in which a node predicts the failure area with the known network coordinates of the failure nodes and determines whether the nodes in their neighbor lists may befailed with a high probability or not by comparing the coordinates of the nodes and thefailure area bounds. Finally, a node can quickly check a node’s state by sending aheartbeat message to the node which has a large probability to be failed. Experimentalresults confirm that NEFAP can detect the failure nodes quickly with less detectionlatency even in the case that a large part of nodes fail in the network compared with theexisting node failure detection methods.The message loss detection and recovery is the process of finding and getting theold messages, which is the key technique to ensure reliable dissemination. To addressthe problem that existing message loss detection and recovery methods require toomuch time and costs for detection recovery, a fast message loss detection and recoverymethod with low redundancy called FDRLR is proposed. FDRLR detects the lostmessages by checking the published records. Moreover, FDRLR employs a BloomFilter to represent the events in the published records and the subscriptions of the nodes.On the one hand, to increase the number of messages carried by the published records toreduce the storage overhead of a single event, so as to ensure the reliability of messageloss detection. On the other hand, a node can quickly determine the lost messages bycomparing the Bloom Filter values in published records and the Bloom Filter values intheir subscriptions. FDRLR recovers the lost messages by backtracking andrepublication. In addition, the lost records of messages are merged in the backtrackingprocess to reduce the communication overhead. The nodes that cache the lost messagesin the backtracking paths republish the lost messages according to a dynamic multicasttree to improve the recovery process. Experimental results confirm that FDRLR detectsthe lost messages faster than the existing message loss detection and recovery methodswith less redundancy.