One of the characteristics of MPLS TE is that an established LSP (label Switched Path) is not torn down in the event of a bandwidth increase process failure (RFC3209). LSP is “out of profile,” so it is not following the traffic engineering parameters that were set up for it. When LSP is out of profile, it may result in degraded performance and traffic loss.
Moreover, this can cause issues when there is an increase in bandwidth demand on one LSP that is sharing links with other in-profile LSPs that are not aware of the demand increase. If the increase in demand cannot be satisfied by other links, the affected LSP may stay on the same path, causing congestion and affecting other in-profile LSPs that share the same link.
Suppose we have a network topology where two routers, R1 and R2, are connected by a single link. Let’s assume we have two LSPs, LSP1 and LSP2, that traverse this link. Both LSPs require a certain amount of bandwidth to operate properly.
Let’s assume that the traffic demand on LSP1 suddenly increases, resulting in a situation where LSP1 requires more bandwidth than what is available on the link between R1 and R2. At the same time, LSP2 is still active and is also using the link.
As a result, LSP1 and LSP2 are now in a deadlock situation. LSP1 is waiting for more bandwidth to become available on the link, but LSP2 is already using the available bandwidth and cannot release it until it has completed its transmission. Similarly, LSP2 cannot proceed until LSP1 releases some of the bandwidth it is using.
In this scenario, LSP1 and LSP2 are deadlocked and cannot make any progress until the deadlock is resolved. If the deadlock persists, both LSPs may fail to transmit data, which can result in severe network performance issues.
In distributed MPLS traffic engineering, we can use a combination of Auto-Bandwidth and LSP Priority to prevent deadlock.