TABLE OF CONTENTS In this chapter we will define GMPLS path computation, and why and when it is required by other sub-systems of the GMPLS control plane. We will show how a transport network can be presented as a weighted, connected graph, and how graph theory can be used to determine efficient paths on to which GMPLS services can be mapped. We will describe some basic popular unconstrained path computation algorithms and, after that, we will show which constraints must be considered in order to determine optimal paths in real networks.
We define a transport service as a way of delivering user traffic of a certain type from the service source point to the service destination point with a certain quality. The user traffic characteristics, as well as the service quality, are agreed upon by the user and the Service Provider. We define a path as a sequence of Service Provider network resources, which, if properly provisioned, can realize the service. Once a path is determined it can be signaled between control elements that manage the selected resources to perform the provisioning (resource allocation, building of the cross-connects, and so on).
Path computation is the process of selecting or determining the path, and can be performed either at the time of, or ahead of service provisioning. The former is called on-line path computation and the latter, off-line path computation. The hybrid case is also possible where some (usually initial) path computation is performed off- line and some is performed...
