A university network will grow from 8 to 28 iBGP routers in one AS with multiple points of presence, and full-mesh peering is becoming operationally heavy with churn and configuration burden. Designers need stable loop-free scaling without splitting the AS or renumbering. Which iBGP scaling design for route reflectors and confederations best fits this requirement?

Hierarchical route reflectors with redundant clusters replace n-squared peering with layered client sessions to support growth from 8 to 28 speakers while preserving one autonomous system and existing addressing, and tuned distinct cluster IDs per site preserve ORIGINATOR_ID and CLUSTER_LIST loop prevention so redundant paths remain available without loops or churn. The confederation design is valid for very large domains because sub-autonomous systems localize mesh and policy, yet it violates the explicit no-split requirement and adds migration and sub-AS path complexity the scenario rejects. The optimized full-mesh design with peer groups and dampening is legitimate for small stable networks and preserves native loop-free behavior without hierarchy, yet it leaves session count, configuration burden, and update replication unscaled at 28 nodes across multiple points of presence. The hierarchical design sharing one cluster ID network-wide is tempting for simpler management and functions inside a single cluster, yet it weakens loop detection and redundancy discrimination across sites, causing excessive suppression and poor path choice that undermines stability. Exam caveat is to separate designs that divide the autonomous system from those that preserve it, since both reduce sessions but only the latter satisfies the stated constraint. Practical check is to audit cluster IDs per location, verify redundant reflector peerings and client hierarchy, and inspect CLUSTER_LIST in show ip bgp before expanding beyond twenty peers.