Publications

Abstract : In this paper, we have presented an efficient traffic grooming algorithm (DTMR) for selecting the optimal path based on the varying traffic load conditions. In this algorithm, the initial incoming packets are sent through all the possible paths. With the monitored values on each path, the source estimates the blocking probabilities of each path and the source node selects the best three paths. Then the traffic is classified into high priority and low priority by the source. Subsequently, they are classified into high speed and low speed traffic. High priority high speed data is sent through primary path. Low priority high speed data is sent through secondary path. The high priority low speed packets are inserted in a frame of fixed length containing the given packets along with replicas of the same packet. Enough replicas are added so that the size of the frame adds upto a minimum of 3.5 Gigabits and a maximum of 5 Gigabits. These high priority low speed packets are sent through the primary path. The same procedure is done for low priority low speed packets as well. These low priority low speed packets are then sent through the secondary path. Backup path is kept for survivability of the network. However, the initial duplication of packets is only done if there are no other requests waiting in queue. When these high priority low speed or low priority low speed packets arrive at another node, and if there are other low speed requests which arrive at this node to be processed; then the replica packets are discarded and the low speed packets at the other node is added in the frame. Using the number of acknowledgement packets that are received, the blocking probabilities of primary path and secondary path are updated. A threshold value thresh is set. Only if the maximum value of the blocking probabilities of primary path or secondary path goes above thresh, the algorithm restarts and identifies and assigns optimal paths, thus optimizing the overall setup delay of the system. In addition the algorithm has a proactive approach, such that it minimizes the chance of blocking and also the approach is self-acting as it automatically changes with the traffic load variation across the network. By simulation result, we show that our algorithm has low blocking probability and high throughput which improves the QoS of the network.