2.6. DEPLOYMENT CONSIDERATIONS
The number of LSPs that are used to implement the traffic engineering requirements is one of the major scalability concerns for MPLS-TE deployments. Two major factors impact the number of LSPs in the network:
The extent of the deployment. For any-to-any connectivity, a solution based on RSVP-TE requires a full mesh of LSPs. Assuming N devices meshed, this yields O(N2) LSPs. For this reason, many MPLS-TE deployments are limited to the core of the network, as explained in the Foundations chapter (Chapter 1). Solutions to this problem using LSP hierarchy are also discussed there.
The size of the reservations. If the size of the traffic trunk between two points exceeds the link capacity, one solution is to set up several LSPs, each with a bandwidth reservation that can be accommodated by the link capacity. Traffic can then be load-balanced between these LSPs. Although logically a single LSP is necessary, the limitation on the maximum size of the reservation causes several LSPs to be set up in this case. An increasingly popular solution to this problem is to use aggregated interfaces, where several physical interfaces are 'bundled' together and treated as a single logical interface from the routing protocol's point of view. Thus, a single LSP can be set up in this case. The downside is that failure of any interface in the bundle causes the LSP not to be able to establish.
In the context of MPLS-TE deployments, vendors typically express ...