The constant expansion of IP VPNs has already led to hierarchical sructures with multiple local/regional IP VPN providers being connected via larger national/global providers. It is particularly challenging to provide essential management services like traffic engineering in such environments. In addition to the well known problems of traffic engineering, the hierarchical environment imposes additional restrictions since different networks belong to different administrative entities and can not share information easily. In this work we present an architecture for supporting traffic engineering across multi-provider hierarchical networks that is based on path pre-computation and advertising appropriately summarized topology information about the parent provider's network into the IGP of the customer networks.
We show that this path pre-computation based approach can indeed fulfill the requirements of hierarchical traffic engineering and has numerous architectural advantages when compared to an on-demand client/server path computation model. In this talk we present this architecture, explore and provide solutions to some of the problems of pre-computation in the traffic engineering context and compare our solutions with on-demand client/server path computation.
George of Redback's Software Engineering team, holds a Ph.D in Computer Sciecne from the university of Maryland, at College Park. He has worked in IBM T. J. Watson Research Lab and since 1999 he is with Redback Networks where he has worked on OSPF, RSVP-TE and L2VPN.