Lazy Rerouting for MPLS Traffic Engineering

We propose novel Traffic Engineering (TE) schemes for congestion control in MPLS networks based on a reactive mechanism. While most existing TE schemes to prevent network congestion rely on constraint-based routing (CBR), the proposed algorithms use a local search technique where the basic move is the modification of the route for a single Label Switched Path (LSP). Because modifications cause a temporary disruption in the network, a "laziness" criterion implies that moves are executed only when absolutely necessary or when the situation is very close to requiring it. Two versions of the algorithm are proposed: in the first one, called FID, an already established LSP is rerouted when a certain level of network congestion is detected, while in the second, called Lazy FID (or LFID), it is rerouted when a new LSP request cannot be satisfied. Experiments under a dynamic traffic scenario show a reduced rejection probability especially with long-lived and bandwidth consuming connection requests, thus proving a better network resource utilization compared to existing CBR schemes in MPLS networks, while guaranteeing a reduced computational complexity.