MPLS Based Best Effort Traffic Engineering

The advent of Multi-protocol Label Switching (MPLS) enables traffic engineering by introducing connection-oriented features of forwarding packets over arbitrary non-shortest paths. Our goal in this research is to improve the network utilization for best effort traffic in IP networks. By examining the best effort traffic class, we assume The large volume of research that has been conducted on traffic engineering for assured forwarding and expedited forwarding with admission control allows traffic engineering on those traffic classes. We examine a different and in some sense a "more complex" problem (traffic demands are not known a priori) of traffic engineering for the remaining network bandwidth which is utilized by best effort traffic. We present a generic traffic engineering framework and four specific algorithms. This framework has two prominent features 1) it uses MPLS to encapsulate source-destination aggregate flows within a Label Switched Path (LSP), with multiple LSPs per source-destination. 2) The framework is "stateless", only the topology is used to determine the traffic routing.