Cisco Cisco IOS Software Release 12.0(23)S

link failures are in practice independent with high probability. This “independent failure assumption” in 
combination with backup tunnels signaled without explicit bandwidth reservation enables efficient 
bandwidth sharing that yields substantial bandwidth savings.

Backup tunnels protecting the sub-pool traffic do now draw bandwidth from any pool. Primary traffic 
using the global pool can use the entire global pool, and primary traffic using the sub-pool can use the 
entire sub-pool. Yet, sub-pool traffic has a complete bandwidth guarantee if there is a single link failure. 

A similar approach can be used for node and SRLG protection. However, the decision of where to put 
the backup tunnels is more complicated because both node and SRLG failures effectively result in the 
simultaneous failure of several links. Therefore, the backup tunnels protecting traffic traversing all 
affected links cannot be computed independently of each other. The backup tunnels protecting groups of 
links corresponding to different failures can still be computed independently of each other, which results 
in similar bandwidth savings. 

Backup-bandwidth is used locally (by the router that is the head-end of the backup tunnel) to determine 
which, and how many, primary LSPs can be rerouted on a particular backup tunnel. The router ensures 
that the combined bandwidth requirement of these LSPs does not exceed the backup-bandwidth.

Therefore, even when the backup tunnel is signaled with zero bandwidth, the backup-bandwidth must be 
configured with the value corresponding to the actual bandwidth requirement of the traffic protected by 
this backup tunnel. Unlike the case when bandwidth requirements of the backup tunnels are explicitly 
signaled, the value of the signaled bandwidth (which is zero) is not the same value as the 
backup-bandwidth.