Path Computation Element – Bringing the WAN to the SDN era

The move towards all-IP networks means that more and more mission-critical applications are being ported to IP networks. The constraints of these applications in terms of bandwidth, latency, jitter, etc. may require the implementation of a differentiated routing policy in the network, whereas the IP network uses a single "shortest" path policy by default. The implementation of traffic engineering techniques based on MPLS (Multi Protocol Label Switching) is often necessary to open up the possibility of calculating constrained paths.

Traffic engineering is not a new concept per se, and was already in use in networks such as ATM (Asynchronous Transfer Mode). It is also deployed to a greater or lesser extent in IP networks to address the need for routing differentiation for different types of traffic.

In this article, we will first review the basic concepts of traffic engineering in an IP/MPLS network, and then focus on the limitations of the distributed approach currently deployed. Secondly, this article introduces the centralized traffic engineering architecture using a PCE (Path Computation Element) to overcome these limitations. The functioning of the communication protocol used by the PCE is detailed, as well as the implementation of a routing architecture using a PCE. This article also presents an analysis of several PCE use cases.

Finally, we look at the safety aspects of introducing PCE, and conclude with a non-exhaustive overview of the current market.