High available method for border gateway protocol version 4
View Patent ↗High availability BGP4 is based on redundant hardware as well as redundant software that replicates the RUN state of BGP4. There are two copies, respectively active and backup, of BGP4 running on two separate redundant hardware platforms. All BGP4 internal implementations apply various methods to replicate the running state of BGP4 independently of peer network routers. When this hardware or software fails on one redundant hardware platform, peer routers are unaware of the failure. Internally, based on duplicative states, the local router recovers from the failure and keeps the protocol running. During the recovery period, the local router can bring up a backup again. In the HA architecture, these activities are not detected by peer routers, such that there is no instability to the Internet backbone caused by BGP4 failure.
1. A method of highly-available Border Gateway Protocol (BGP) routing in a network, comprising the steps of:
establishing BGP peer router connections;
exchanging routing information with said BGP peer routers;
running BGP protocol on two redundant physically separated control plane master control units, such that one is an active BGP instance and the other is a backup BGP instance by bringing up said active BGP instance before said backup BGP instance;
synchronizing the running configuration of said backup BGP instance with the running configuration of said active BGP instance using message transmission from said active instance to said backup instance through a highly reliable Transmission Control Protocol (TCP) link;
copying routing information from said BGP peer routers onto said backup BGP instance;
processing said routing information at said backup BGP instance, such that said backup BGP instance does not advertise said routing information;
maintaining dynamic state synchronization of said backup BGP instance with said active BGP instance, such that said active BGP instance enters an ACTIVE-PROTECTED state and said backup BGP instance enters a BACKUP-PROTECT state; and
in the event of fail-over of said active BGP instance, then seamlessly recovering without detection of said fail-over by said BGP peer routers in said network, by functionally substituting said BACKUP-PROTECT backup BGP instance for said ACTIVE-PROTECTED active BGP instance, such that said BACKUP-PROTECT backup BGP instance establishes itself as a new active BGP instance.
2. The method of claim 1 wherein said step of exchanging comprises selecting and advertising of best routes by said active BGP instance.
3. The method of claim 1 wherein said step of exchanging utilizes a TCP socket to represent each BGP peer router connection.
4. The method of claim 3 further comprising cloning said TCP socket onto said backup BGP instance.
5. The method of claim 4 wherein said step of copying comprises reading by said backup BGP instance from said cloned sockets.
6. The method of claim 1 wherein in the event of failure of said backup BGP instance, then said active BGP instance continues to function seamlessly without detection of said failure by said BGP peer routers in said network.
7. A method of highly-available Border Gateway Protocol (BGP) routing in a network, comprising the steps of:
establishing BGP peer router connections;
exchanging routing information with said BGP peer routers;
running BGP protocol on one of two redundant physically separated control plane master control units, such that said one control unit is an active BGP instance;
establishing BGP peer connections with said active BGP instance and exchanging routing information with the peers;
after the active BGP instance is brought up, bringing up the other said master control unit as a backup BGP instance;
in any order, establishing a highly reliable Transmission Control Protocol (TCP) link between said active BGP instance and said backup BGP instance;
after said backup BGP instance is brought up and after said TCP link has been established, said backup BGP instance signaling said active BGP instance to show its presence, but not signaling said BGP peer routers;
synchronizing backup BGP instance with routing information from said active BGP instance and then cloning onto backup BGP instance a TCP socket that represents each peer connection;
processing said routing information at said backup BGP instance such that said backup BGP instance does not advertise said routing information;
after the cloning step is complete, both said active and said backup BGP instances begin reading from the cloned sockets to learn routes from peer routers, but only said active BGP instance advertising anything to any peers.
8. The method of claim 7 and after said cloning step is complete, maintaining dynamic state synchronization of said backup BGP instance with said active BGP instance, such that said active BGP instance enters an ACTIVE-PROTECTED state and said backup BGP instance enters a BACKUP-PROTECT state; and
in the event of fail-over of said active BGP instance, then seamlessly recovering without detection of said fail-over by said BGP peer routers in said network, by functionally substituting said BACKUP-PROTECT backup BGP instance for said ACTIVE-PROTECTED active BGP instance, such that said BACKUP-PROTECT backup BGP instance establishes itself as a new active BGP instance.