LSP Preemption

When there is not enough unallocated bandwidth on an interface to fulfill the requirements of a new LSP that has passed admission control, existing LSPs that have a lower priority may be preempted. When preemption occurs, bandwidth allocated to lower-priority LSPs is reallocated to the higher-priority LSP. LSP preemption depends on the bandwidth requirements and priority of the new LSP, compared to the bandwidth allocation and priority of already existing LSPs.

When LSP preemption is necessary, the device uses the following rules:

Note

LSP preemption rules have changed to improve the scalability and performance for Fast Reroute (FRR) enabled LSPs. See bullets three and four below for changes to LSP preemption for FRR enabled LSPs.

Preempt existing LSPs that have lower priority than the new LSP
When several existing LSPs have lower priority than the new LSP, preempt the LSP that has the lowest priority
When two LSPs have equal priority and one LSP must be preempted, preempt the LSP which is currently FRR enabled irrespective of its bandwidth requirement
Preempt as many FRR enabled LSPs as necessary before preempting unprotected LSPs of the same priority. For example, when both FRR enabled LSPs and non-FRR enabled LSPs are configured, the system attempts its best to preempt FRR enabled LSPs first before preempting non-FRR enabled LSPs until the bandwidth requirement is met for a new high priority LSP

In the example above, bandwidth has been allocated to an LSP that has a hold priority of three and a mean-rate of 1,000 Kbps. When a new LSP with a setup priority of two, hold priority of one, and mean-rate of 10,000 Kbps is established, admission control, bandwidth allocation, and LSP preemption work as described below.

Admission control: On the interface, there is 10,000 Kbps available to priority two. The mean-rate for the new LSP is 10,000, so the LSP passes admission control; bandwidth can be allocated to it.
Bandwidth allocation: The hold priority for the new LSP is one. On the interface, 10,000 Kbps is available to priority one. This entire amount is allocated to the LSP.

LSP preemption: The first LSP had been using 1,000 Kbps of this amount, but its hold priority is only three. Consequently, the first LSP is preempted, and its bandwidth allocation removed in order to make room for the new LSP.

Once this happens, the reservable bandwidth array for the interface looks like this:

Table 1. LSP preemption bandwidth allocations
Priority	Unreserved Bandwidth
0	10,000
1	0
2	0
3	0
4	0
5	0
6	0
7	0
Active : LSP with setup 2, hold 1, mean-rate 10,000 Preempted: LSP with setup 6, hold 3, mean-rate 1,000

On this interface, the only LSP that could preempt the active LSP would be have a setup and hold priority of zero.

When multiple LSPs are candidates for preemption, the device normally preempts the LSP with the lowest priority. However, when preempting a higher priority LSP with a high bandwidth requirement would allow lower priority LSPs with lower bandwidth requirements to avoid preemption, the higher-priority LSP is preempted.

For example, consider an interface with 10,000 Kbps of reservable bandwidth, allocated to two active LSPs: one with a setup priority of three, hold priority of two, and mean-rate of 5,000 Kbps; and another with a setup priority of four, hold priority of three, and mean-rate of 2,500 Kbps. When an LSP with a setup priority of one, hold priority of zero, and mean-rate of 7,500 Kbps is established, the following take place.