Back Recursive Path Computation

A multi-domain network presents a mayor obstacle to the effective computation of an end-to-end path that resides on the fact that a multi-domain network contains various independently managed system (autonomous system) that have confidential information on their own networks topology and state mainly due to security reasons. 

This information confidentiality prevents each single domain PCE to have full topology and Traffic Engineering information on the whole network which is necessary to compute an optimal or even feasible end-to-end path. 

In order to achieve a feasible end-to-end path computation it is possible to implement distributed or hierarchical architectures relying on the PCE. One possibility is to compute the path in a distributed way where each domain PCE computes the LSP for its own domain. In this architecture, usually, in domain sequence to be crossed is previously determined (normally via the BGP protocol). 

In a per domain path calculation, entry border nodes of each domain request its domain PCE for a LSP inside that domain, and therefore the resulting end-to-end path is a concatenation of those LSPs. In Back Recursive Path Computation (BRPC) PCEs present in each domain of the sequence collaborate to create a Virtual Shortest Path Tree. This VSPT comprises all possible end-to-end paths that cross the stipulated domain sequence. A request message is sent to the PCE in the destination domain and this one sends a VSPT to the PCE in his neighbor domain belonging to the sequence and so on so forth until the all the VSPT from all domains in the sequence reach the PCE in the source domain. Once the VSPT is informed to the source domain PCE, this selects the final path (Vasseur et al.,2009).

Vasseur, J.P. et al., “A Backward-Recursive PCE-Based Computation Procedure (BRPC) to Compute Shortest Constrained Inter-domain Traffic Engineering LSPs,” IETF RFC 5441, Apr. 2009.