IP networks were not designed to repair failures in sub-second intervals, but applications such as VoIP increasingly are driving the need for rapid failure detection and correction. Traditional routing infrastructure has been limited in meeting the failure-resolution requirements of real-time applications such as voice. A new protocol, Bidirectional Forwarding Detection (BFD), is helping to overcome these limitations and increase the speed of failure detection and recovery. An IETF draft standard, BFD provides a simple, lightweight and abstract method of detecting the ability of links or systems to forward traffic.
BFD is sufficiently abstracted from underlying transport technologies so that it can detect failures at many layers. It can be used to monitor the validity of Ethernet networks, Multi-protocol Label Switching (MPLS) label-switched paths, Generic Routing Encapsulation or IPSec tunnels, or virtually any other type of transport.
At its heart, BFD is a high-speed stand-alone hello protocol (similar to those used in routing protocols such as Open Shortest Path First or Intermediate System-to-Intermediate System that can be associated with a link, interface, tunnel, route or other network forwarding component).
BFD neighbour systems negotiate a peer relationship, and each monitors the flow of BFD packets coming from the other systems at the negotiated rate. This can be specified in sub-millisecond increments.
When a peer system misses receipt of a certain preconfigured number of packets, it infers the failure of the BFD-protected software or hardware infrastructure, whether it be a label-switched path, a tunnel of some other type or a switched Ethernet network.
BFD is implemented in the control plane of routers and other systems. A network failure detected by BFD can be corrected by the forwarding plane (for instance, in MPLS fast reroute), or by the control plane (for example, when BFD is used to speed up the operation of routing protocols).
The simplicity of BFD makes it possible to use it in some applications that have never had an adequate solution for forwarding failure detection. Consider, for example, an application wherein a VoIP media gateway is connected to an IP core over a switched Ethernet network.
This type of application traditionally presented two challenges: currently there is no network-visible means of detecting failures between it and the IP edge router; and Ethernet has no way of notifying a host or router that a failure has occurred in a distant part of a switched network. Rapid link failure detection is important for high availability of VoIP networks, but it is not easy to rapidly detect link failure when an intervening switch is present between host and router.
Media gateway component
BFD is simple enough to be included in media gateway platforms and in this example can be used to maintain connectivity between the gateway and its edge routers.
Failure of the intervening Ether-net segment or segments will be detected by BFD, which lets the gateway and router switch to alternate redundant paths. Once a failure is detected and verified, BFD can trigger the resolution mechanisms in all routing, transport and tunneling systems.
Because its mission is simple and abstract, BFD can concentrate on finding forwarding faults as quickly as possible, and thereby let voice, video and other demanding services get the forwarding service they need to be successful. Protocols such as BFD are letting service providers deliver VoIP and other real-time services over IP with the levels of reliability and availability customers demand.
Kolon is senior technical solutions manager for Juniper Networks Inc. He can be reached at mkolon @juniper.net.