A New Preemption Avoidance and Load Balancing Policy for Traffic Engineering in Mpls Networks

The objective of traffic engineering is to optimize resource efficiency and network performance. Its study issue is to make the use of the available bandwidth in IP backbone networks effectively. MPLS (Multi-Protocol Label Switching) approach proposed by the IETF (Internet Engineering Task Force) is a new technology that facilitates significantly the process of traffic engineering. In MPLS networks, the higher-priority LSP (Label Switching Path) will preempt the resource of lower-priority LSP when the bandwidth resource is strained. The LSP preemption introduces a setup and holding priority. When preemption occurs, the lowerpriority LSP will be destroyed, and its bandwidth resource is released. The higher-priority LSP obtains the freed bandwidth resource to establish its path, while the preempted lower-priority LSP has to be rerouted by selecting another LSP. However, the LSP cannot ensure that bandwidth resources won’t be preempted again. If this situation occurred frequently, routers would have superfluous overhead and encounters an awful quality of service. In this paper, a new policy for preemption avoidance and load balancing is proposed. It complements with traditional Constraint-based Routed (CR) scheme and aims to network resource utilization and the number of LSP preemption. The difference is considered between the traditional method and our policy that depends on the preemption probability of the LSP. The preemption probability combines two criteria: the priority of LSP to be delivered, and the remaining bandwidth of link along the LSP. Each node of LSPs calculates the preemption probability, the minimal preemption probability LSP is selected to service the flow. The simulation results show that our policy with comparable network performance is better than CR.