A class-based adaptive congestion control mechanism in MPLS node

In this paper, we demonstrate a mechanism for handling the congestion in MPLS LSR node. Based on the analysis on the output queues, we develop a class-based virtual queue model. By adjusting the number of unit FIFO allocated to virtual queues of multiple classes, self-adaptation for the traffic burst that may lead to congestion is achieved, which results into packet loss reduction and meets the requirement of QoS. The algorithm of expanding the bandwidth to solve congestion as the traffic burst occurs for one class is specified in this paper. Also based on the analysis of traffic for one class, the rule for handling the burst of multiple traffic is given as well. In addition to this, the ANGBA algorithm is applied here for self-adaptive bandwidth adjustment.