Hi Wang, I read the draft. It is interesting but one point isn't clear for me. Why couldn't we extend routing protocol to make border nodes get the bidirectional inter-AS links?
Thanks, Frederick > > Hi JP and PCEers, > > Thank you for the promotion.We would like to trigger the > discussion, and greatly appreciate your comments and participatory. > > We add some scenarios to describe why we need to extend BRPC > procedure in the case of computing inter-AS bi-directional path. > > Suppose the inter-AS link fails in one direction(e.g., from C1 to B1). > As border node just have unidirectional inter-AS TE link properties, > for a bi-direction path computation from A to Z, PCE1 doesn't know > that the link from C1 to B1 can not be used. How PCEs select the > inter-AS links for the bidirectional path? > > Similarly,Suppose uni-direction LSP along A1-C-B-Z1 have already been setup. > After that we setup a bi-direction LSP from A to Z, neither PCE1 nor > PCE2 knows that whether the inter-AS link can > satisfy the constraints in both of the directions or not. > > For a bi-direction path computation from A to Z : > PCE2 computes bi-direction optimal paths(i.e.,VSPT): EGZ and FZ. > > Suppose the uni-direction inter-AS link from E to C doesn't satisfy > the constraints. Only inter-AS links C-F,D-E and D-Fcan > satisfy the constraints in bi-direction. But neither PCE1 nor PCE2 > knows that information, they don't know which inter-AS > links should be selected. > Actually the optimal paths between exit BN of AS1 and node Z are > shown in red color bellow, but PCE2 cannot compute > these bi-direction optimal paths by adding the inter-AS links . On > the other hand, PCE1 cannot compute the > bi-direction optimal paths by adding the inter-AS links either (i. > e.,PCE1 can not compute the bi-direction path from A to E and A to F). > > > > Thanks . > > Xuerong Wang / Best Regards
_______________________________________________ Pce mailing list [email protected] https://www.ietf.org/mailman/listinfo/pce
