An enterprise network has two core routers that connect to 200 distribution routers and uses full- mesh iBGP peering between these routers as its routing method. The distribution routers are…

Adding memory does not reduce the number of BGP sessions or the CPU cycles required to process BGP updates across a full-mesh topology.

The core routers are not identified as the bottleneck - distribution routers are experiencing high CPU, so upgrading core router memory does not address the problem.

In a full-mesh iBGP topology with n routers, each router requires n-1 peering sessions, resulting in n(n-1)/2 total sessions. With 202 routers, this is over 20,000 sessions. Implementing route reflectors on the two core routers allows distribution routers to peer only with the reflectors instead of every other router, reducing peering sessions from hundreds to just two per distribution router, which directly resolves the CPU issue.

Bandwidth between core routers is unrelated to the BGP session count and processing overhead on distribution routers.

eBGP is used between autonomous systems, not within one AS - using eBGP between core and distribution routers would change the AS design fundamentally and is not the correct fix for full-mesh scaling.