Extending MPLS Traffic Engineering to deal with QoS

Traffic Engineering (TE) deals with the performance optimization in operational networks, aiming for example at the fair distribution of traffic in order to avoid or minimize congestion. Multi Protocol Label Switching (MPLS) is a key technology that enables advanced TE functionality. The MPLS framework consists of several components, such as the label switched forwarding plane, the label distribution and routing protocols and the “routing decision engine”. Extensions to these components to fully support Traffic Engineering are currently under discussion within standardization bodies. Combining the proposed extensions to different control plane components in a consistent architecture is not a trivial operation. This paper defines the architecture of a Traffic Engineered MPLS network and describes its implementation in a test-bed composed of Linux PCs acting as MPLS routers. The architecture is aligned with current discussion within IETF and includes as an option an extension to RSVP-TE proposed by the authors. The prototype implementation allows verifying the correct functional behavior of control plane protocols. The test-bed provides a flexible platform where further extensions can be tested and some kinds of performance tests are possible.