Seamless Detection of Link and Node Failures in Label Switched Networks with Local Restoration

We consider a label switched network in which backup paths are provided using a local restoration scheme [1], [2]. A node receives a failure notification if either an adjacent link or a neighboring node fails. Such a node, however, cannot distinguish between link and node failures and must do one of two things. 1) Activate backup paths corresponding to both the link and the node suspected to have failed, or 2) Employ a mechanism to establish whether it is a link or a node that has failed, and activate the requisite backup paths once the failure event has been identified. Incase backup paths are activated without disambiguating between link and node failure, bandwidth sharing estimates for a backup path must be revised to take into account concurrent activation of certain additional backup paths. Consequently, a greater amount of bandwidth has to be provisioned when the failure detection is ambiguous. In the case wherein a node waits to identify the exact failure before activating the requisite backup paths, there is increased switchover latency subsequent to the network failure. Evidently, the increased switchover latency translates into a greater traffic disruption following a network failure. We present a simple solution to this problem. Our solution eliminates the need to over-provision backup bandwidth. Moreover, our solution makes possible immediate activation of backup paths without waiting to disambiguate between link and node failure. The key idea is that if an intermediate node along the activated backup paths encounters a resource reservation violation, it can infer the exact type of failure that has transpired. It may then abort the traffic corresponding to the erroneously activated paths while the network traffic that requires restoration can flow through without being disrupted.