Redundant LSPs

There are three methods for provisioning redundant RSVP-TE LSPs at the ingress LER, also referred to as head-end LSP protection:

Secondary RSVP-TE LSPs can be configured to provide backup LSPs in the event that the primary LSP fails. You can create up to two secondary LSPs for each primary LSP. The secondary LSPs are fully provisioned, pre-established RSVP-TE LSPs that are maintained as inactive until needed. If the primary LSP is torn down, the associated LSP next hop is removed from the route table, and a new LSP next hop representing one of the secondary LSPs is installed as the preferred LSP. If there are multiple secondary LSPs available, the secondary LSP is randomly selected. If the primary LSP is re-established, the primary LSP next hop information is re-installed and the secondary LSP returns to inactive state.

If both the primary and secondary paths for an LSP fail, and there are no other RSVP-TE LSPs active to the destination, an LDP LSP can be used if available.

Operation with L2 VPNs is similar. If a primary path fails, and a secondary LSP is available, VPLS uses the secondary LSP. When the primary LSP is re-established, VPLS again uses the primary LSP.

Specifying redundant LSPs is accomplished by assigning secondary paths to an LSP. The configure mpls rsvp-te lsp lsp_name add path path_name secondary command can configure the specified path as a backup LSP. A path different from the primary path must be specified. It is recommended that defined paths be configured using EROs to specify different paths through the network. Relying on the routing topology, by configuring the path to any, can create two LSPs that take the same path. It is important to understand that the configured LSP signals multiple LSPs, up to three (one primary and two secondary), but only one LSP can be used to forward traffic at any one time.

Fast Reroute LSPs are based on the on IETF RFC 4090, Fast Reroute Extensions to RSVP-TE for LSP Tunnels, which defines RSVP-TE extensions to establish backup LSP tunnels for local repair of LSP tunnels. To respond to failures, these mechanisms enable the re-direction of traffic onto backup LSP tunnels in tens of milliseconds, and this meets the needs of real-time applications such as voice over IP (VoIP). This timing requirement is satisfied by computing and signaling backup LSP tunnels in advance of a failure and by re-directing traffic as close to the failure point as possible. In this way the time for redirection includes no path computation and no signaling delays, which include delays to propagate failure notification between label-switched routers (LSRs). Speed of repair is the primary advantage of using fast-reroute backup methods.

There are two backup methods; the detour LSP method (which is also called the one-to-one backup method) and the facility backup method (which is also called the by-pass tunnel method). The software supports only the detour LSP method.

Based on the RFC-4090 there are two different methods to uniquely identify a backup path:
  1. Path-specific method
  2. Sender template-specific method

The software supports only the path-specific method, which uses a new object, the DETOUR object, to distinguish between PATH messages for a backup path and the protected LSP.

The following figure illustrates the terminology used to describe fast-reroute configuration and operation.

expand icon
Fast-Reroute Terminology
../Graphics/EX_mpls_0007.svg

The primary LSP in the following figure is established between MPLS routers A and D. Router A is the ingress LER, and Router D is the egress LER. When used with fast-reroute protection, the primary LSP is also called the protected LSP, as it is protected by the detour LSP created by the fast-reroute feature. The detour LSP provides a route around a protected component. In the following figure, the link between Router B and Router C has failed. The detour LSP, which is indicated by the dashed line, provides a path around the failure.

Routers B and C are transit LSRs for the primary LSP. With respect to a specific LSP, any router that is not the ingress or egress LER is a transit LSR. Routers F and G are transit LSRs for the detour LSP.

The origin of the detour LSP is called the Point of Local Repair (PLR), and the termination of the detour LSP is called the Merge Point. A protected LSP is an explicitly-routed LSP that is provided with protection. A detour LSP is also an explicitly-routed LSP. If you configure a series of one or more hops (EROs), then based on the currently set DYNAMIC_FULL option in the Constrained-based Shorted Path First (CSPF) routing component, the CSPF will calculate and try to fill in the gaps to build a complete list of EROs.

You can configure up to two secondary LSPs for each standard TE (non-FRR) LSP or for each protected FRR LSP. If a standard TE LSP fails, then one of the secondary LSPs becomes active. If that secondary LSP fails, the other secondary LSP becomes active. If a protected FRR LSP fails, its detour LSP becomes active. If the detour LSP fails, then one of the secondary LSPs becomes active, and if that secondary LSP fails, the other secondary LSP becomes active. If all configured backup and secondary paths for an LSP fail, a different active RSVP-TE LSP to the destination can be used. Otherwise, an LDP LSP can be used if available.

The primary advantage of detour LSPs is the repair speed. The cost of detour LSPs is resources. Each backup LSP reserves resources that cannot be used by other LSPs. Another cost is that currently there is no automatic way to redirect traffic from a detour LSP back to a primary LSP when the protected LSP recovers. Redirecting traffic from the detour LSP to the primary LSP requires a series of CLI commands.

Fast reroute protection is configured primarily on the ingress LER, however, it must be enabled on all transit LSRs and the egress LER also. After configuration is complete and fast-reroute protection is enabled on the primary LSP, the primary and detour LSPs are signalled. Provided that the resources are available, detour LSPs are set up at each transit LSP along the primary LSP.

Multiple RSVP-TE LSPs can exist or be configured to the same destination. The paths do not need to be equal cost; all that is required is that all the LSPs to the same destination must have IP transport enabled. In this scenario, LSP next hop information is communicated to the route table for up to eight different named RSVP-TE LSPs. Locally originated traffic is distributed across each LSP based on standard IP address hash algorithms. If one of the LSPs fails, the traffic is redistributed across the remaining active named LSPs. Unlike the backup LSP mechanism, all of the redundant multipath LSPs are unique named LSPs and in general have primary configured paths.