Load Balancing Algorithm Based On Adaptive Number Of Message Copies In Delay Tolerant Network

Delay/disruption tolerant networks which is proposed as a newly end-to-end,store-and-forward network architecture are characterized by intermittent connectivity,extremely high latency, high error rates, heterogeneous interconnection, etc.Traditional routing protocols for internet is difficult to fulfill its needs. Recent years,routing algorithm based on flood policy have been attracted most researchers’attention because of the high delivery probability in many proposed algorithm, butflood policy which rely on increasing message copies to improve transmit efficiencyhave its own disadvantage. It easily lead to network congestion, and fierce resourcescompetition between nodes. Therefore, we propose a load balancing algorithm whichadaptive message copies, The main contribution of this algorithm as follows:(1) With the growth of the network operation time, the traditional floodingalgorithm will lead to network congestion due to excessive replication of messagecopies, and no congestion control mechanism can make a reasonable response basedon network conditions. This paper presents a new regional congestion metric--congestion degree of cluster, when it is more than the network threshold, restrictedgrowth of message copies by multiplicative less the initial value.(2) Some nodes in the network because of its high activity, usually been selectedas the next hop that formed a focal point of the message. But its energy and cacheresources are limited, frequently been selected as the next hop of the messageincreased the burden of the node itself, which made the more intense competition forresources within the node. In this paper, a node-level congestion metric namedcongestion degree of node have been proposed, when selected the next hop formessages, the current node sort all connections by node congestion degree and selectthe minimum congestion degree of the node to transmit in order to avoid loadunbalance in the network.(3) Traditional flooding algorithm delete the package is usually based on thesurvival time of the message, the delete mechanism has certain blindness, which cannot correctly estimate the value of the packet. In this paper, we sorted messages bypriority in the node cache, when delete the package, removed the package with thesmallest priority, greed ensure that the messages have maximum value in the nodecache.Experiments shows that presented load balancing algorithms have betterperformance than the traditional flood algorithms. And on the same conditions,comparing with traditional Spray-and-wait algorithm, the algorithm reduces the overhead of the network by about50%, almost66.7%package loss has gone down.