13.4.3 BGP4 Multipathing
BGP4 multipath routing enables traffic to be distributed equally among multiple paths (routes) to one destination network. The following describes how BGP4 multipathing operates in the Switch.
- <Structure of this section>
(1) Multipathing BGP4 Routes by Multipathing IGP Routes
In the Switch, the next hop of a BGP4 route is resolved through an IGP route. If there are multiple IGP routes corresponding to the NEXT_HOP attribute value of the BGP4 route being resolved, the BGP4 route will also take multiple paths. The following figure illustrates the concept of multipath generation.
|
In this figure, the routers are connected by two physical interfaces. For peering, they are configured to use the address assigned to the peer router itself. You can assign a peer address to each router using the ip address configuration command with the loopback interface specified. Also, you can specify use of the device address as the peering address on the local side by setting the neighbor update-source configuration command. If the local side connects to an external peer or member AS peer in this command, you must also set the neighbor ebgp-multihop configuration command.
The BGP4 route (destination network: W, next hop: A) that is reported to Switch 1 by AS100 references an IGP route to resolve the next hop. Next hop: Because the gateways of the IGP route to next hop A are a and b, the gateways of the BGP4 route are also a and b. Similarly, for the BGP4 route (destination network: W, next hop: B) that is reported to Switch 2 by Switch 1, the gateways of the IGP route to next hop B are c and d, so the gateways of the BGP4 route are also c and d.
- Notes on BGP4 multipath routing via multiple IGP routes
-
Only static routes and OSPF routes can be configured as a multipath IGP route in the Switch. For details about the concept of multipathing static routes, see 9.1 Description. For details about the concept of multipathing OSPF routes, see 11.1.7 Equal-cost multipathing.
(2) Multipathing of BGP4 routes learned from two or more peers
In the Switch, you can generate a BGP4 multipath based on equal-cost routes to the same destination, learned from different peers connected to the same neighboring AS. To generate a BGP4 multipath, use the maximum-paths configuration command. By specifying the all-as parameter in this command, you can generate a BGP4 multipath from routes learned from different neighboring ASs. The following table describes the tie-breaker conditions.
Conditions |
Remarks |
---|---|
Equal weights |
- |
Equal LOCAL_PREF attribute values |
- |
Equal number of ASs in the AS_PATH attribute |
The AS_SET path type in the AS_PATH attribute is counted as a single AS. |
Equal ORIGIN attribute values |
- |
Equal MED attribute values |
The tie-breaker based on MED attribute values applies only to redundant routes learned from the same neighboring AS. To include redundant routes learned from different neighboring ASs, specify the bgp always-compare-med configuration command. |
Learned from the same peer type (external peer, member AS peer, or internal peer) |
- |
Equal next hop (equal IGP metric used at next-hop resolution) |
- |
Legend:-: Not applicable
The following figure illustrates the concept of BGP4 multipaths learned from multiple peers.
|
In this figure, Routers 2 and 3 in AS100 notify the Switch of two BGP4 routes to network W, via next hops a and b, respectively. If the routes are in a tie-break state, the Switch will generate gateways based on each route's NEXT_HOP attribute (gateways a and b). When propagating these routes to other BGP4 peers, the Switch advertises the route that has higher priority.