Cisco Technology, Inc.

Junaid Israr

Rakesh Gandhi

Robert Sawaya

Tarek Saad

Eric Osborne

Reshad Rahman

Prabhu Vaithilingam

Anca Zamfir

In one embodiment, certain SRLGs associated with members of a bundle are deemed to be 'risk-free SRLGs', i.e., SRLGs whose failure may not substantially impact the capacity of the bundle to carry traffic in a computer network. In a bandwidth unaware embodiment a number of active members of the bundle is identified for each SRLG, and a bandwidth aware embodiment further specifies an available bandwidth capacity for each SRLG to determine risk-free SRLGs. A backup path or tunnel may be established and utilized to protect a communication link as long as the SRLGs shared are deemed to be risk-free SRLGs.

System and method for link protection using shared SRLG association

In an embodiment, a method is disclosed for minimizing soft preemptions of LSPs. Upon receiving a reservation message for an LSP whose requested bandwidth that exceeds the available bandwidth of downstream links, a network node may select a set of LSPs for soft preemption and share the selection with other nodes along their paths, both upstream and downstream. By coordinating the selection of LSPs to soft-preempt among nodes on the path, fewer LSPs may require soft preemption, which may result in minimizing excessive network disruptions, and thus, allowing the network to function more efficiently.

Soft preemption for minimizing preemptions in a network

An RSVP node that has crashed and restarted sends a non-zero Recovery Time value in a Restart_Cap object of a Hello message. Other nodes receiving such messages can detect, based on the non-zero Recovery Time value, that the sending node is restarting. In particular, a first RSVP node can detect whether a second, neighbor node has restarted regardless of the order of restart with respect to the first and second nodes. As a result, the first node can determine whether to forward PATH messages with Recovery Label or Suggested Label, as appropriate or necessary to rebuild label-switched paths with crashed and restarted nodes. Therefore, when multiple RSVP nodes crash relatively concurrently, the multiple nodes can restart gracefully and automatically detect what kind of communication to use in order to rebuild label-switched paths among themselves.

Method and apparatus for restarting RSVP processes in multiple network devices

Growing community of inventors

Kanata, Canada

Junaid Israr