Abstract: A system of hitless restart in a network, where at least one node in the network provides routing control distributed among ingress ports (ingress cards) and egress ports (egress cards), is disclosed. With distributed routing control, each ingress card has its own routing-control software and each egress card has its own routing-control software. When the routing-software at an ingress port or an egress port of a node is restarted, current connections traversing a restarting ingress card or a restarting egress card continue to function normally during a restart period without data loss. The disclosed system is tailored to a multi-protocol label switching (MPLS) network employing distributed-resource-reservation-protocol traffic engineering (RSVP-TE). The system relies on messaging between ingress card control planes, ingress card data planes, egress card control planes, and egress card data planes of a restarting node.

Abstract: The present invention defines a system and methods for distributed RSVP-TE (resource reservation protocol-traffic engineering) hitless graceful restart for a MPLS (multi-protocol label switching) network. The system comprises a plurality of ingress and egress cards, each card having an MPLS control plane forwarding table for reverse and forward traffic outgoing and incoming labels for LSPs (Label Switched Paths) in the MPLS network; the cards having data planes, each card data plane having said forwarding table stored thereon; and a means for providing messaging between the ingress card MPLS control plane, ingress card data plane, egress card MPLS control plane, and egress card data plane. The methods of the embodiments of this invention do not require hitless graceful restart in telecommunications networks with no requirement for GMPLS stack and therefore can be used in both traditional and generalized MPLS networks.