Traffic Engineering in the Internet: A Survey of Load Balanced Routing

The use of traffic engineering to balance the traffic workload between source and destination helps address the pitfall of the destination-based shortest-path routing paradigm currently used in the Internet. Routing is a primary mechanism for load balancing, and when being employed in conjunction with other QoS mechanisms, scalable QoS as well as efficient resource utilization can be accomplished. The explicite route feature in MPLS has greatly facilitated effective traffic engineering by allowing the source to have a complete control over which paths the traffic will traverse to the destination. As such, parallel paths can be used to distribute traffic load of each traffic class such that the network resources are utilized on both least-cost path(s) and non-least-cost paths to meet constraint and requirements specific to the classes. In this paper, we survey related work in the literature on load balancing in IP networks and MPLS/differentiated service networks, focusing on routing control algorithms. Currently, traffic between same ingress-egress pairs will follow the same path, and therefore the per-node differentiation implies the end-to-end differentiation across the domain. We identify the unresolved problem for future research directions.