IP Route Sharing and ECMP

IP route sharing allows a switch to communicate with a destination through multiple equal-cost routes. In OSPF, BGP, and IS-IS, this capability is referred to as ECMP routing.

Without IP route sharing, each IP route entry in the routing tables lists a destination subnet and the next-hop gateway that provides the best path to that subnet. Every time a packet is forwarded to a particular destination, it uses the same next-hop gateway.

With IP route sharing, an additional ECMP table lists up to 2, 4, 8, 16, 32, or 64 next-hop gateways (depending on the platform and feature configuration) for each route in the routing tables. When multiple next-hop gateways lead to the same destination, the switch can use any of those gateways for packet forwarding. IP route sharing provides route redundancy and can provide better throughput when routes are overloaded.

The gateways in the ECMP table can be defined with static routes (up to 64 way), or they can be learned through the OSPF, BGP, or IS-IS protocols (64-way for OSPFv2, OSPFv3, and BGP; and 8-way for IS-IS). For more information on the ECMP table, see ECMP Hardware Table.


BGP does not use ECMP by default, so if you require that functionality you must explicitly issue the command configure bgp maximum-paths max-paths with a value greater than 1.