RSVP-TE TE++ Dynamic LSP Splitting

RSVP LSPs with the auto-bandwidth feature are increasingly deployed in networks to meet traffic engineering needs. However, the current traffic engineering solutions for point-to-point LSPs are inefficient in terms of network bandwidth utilization, mainly because the ingress routers originating the RSVP LSPs either try to fit the LSPs along a particular path without creating parallel LSPs, or do not interact with the other routers in the network and probe for additional available bandwidth. This feature enhances the RSVP-TE auto-bandwidth feature, whereby an ingress router can acquire as much network bandwidth as possible by creating parallel LSPs dynamically (TE++ dynamic LSP splitting) when RSV-TE auto-bandwidth thresholds are crossed. The flexibility to control the LSP paths is expressed in terms of the configuration for point-to-point LSPs and container LSPs. Controlling the LSP paths using the configuration parameters can be applied under two different aspects: * Topology-There are no topology constraints with this feature. Each member LSP is treated like a point-to-point LSP and is re-optimized individually. An ingress router does not try to compute equal IGP cost paths for all its LSPs, but instead it computes paths for all the LSPs using current traffic engineering database information. While computing a path, constraint-based routing adheres to any constraints specified through the configuration, although there is no change in the constraint-based routing method for path computation. * When to create a new LSP-You can explicitly specify when to create a new LSP. By default, an ingress router periodically computes the aggregate traffic rate by adding up the traffic rate of all the individual LSPs. Looking at the aggregate bandwidth and configuration, the ingress router recomputes the number of LSPs and the bandwidths of the LSPs. The new LSPs are then signaled or the existing LSPs are re-signaled with the updated bandwidth. Instead of looking at the instantaneous aggregate rate, the ingress routers can compute an average (of aggregates) over some duration by removing outlier samples (of aggregates). Managing the LSPs that remain outstanding and active by considering aggregate bandwidth is more scalable than creating the new LSPs based on the usage of a particular LSP. The intervals and thresholds can be configured to track the aggregate traffic and trigger adjustment. These dynamic LSPs co-exist and interoperate with per-LSP autobandwidth configuration. Juniper devices support the following container LSP features: * Equal-bandwidth-based LSP splitting mechanism * Aggregate-bandwidth-based LSP splitting and merging in a make-before-break way * LSP-number-based naming mechanism for dynamically created member LSPs * Periodic sampling mechanisms to estimate aggregate bandwidth * Interoperability with auto-bandwidth feature * ECMP using the dynamically created LSPs * LDP-tunneling on the dynamically created LSP * Static route with container LSP as the next-hop * Configuring container LSP using IGP shortcuts * Aggregated Ethernet links * Logical systems