The operations team debates installing heavy, complex FBF policies on every leaf versus steering at a smaller set of centralized choke points. What placement tradeoff should drive the decision?

When the operations team debates installing heavy, complex FBF policies on every leaf versus steering at a smaller set of centralized choke points, the placement tradeoff that should drive the decision is that distributed FBF on every leaf can increase scale and consistency burden, while centralized steering simplifies policy but concentrates path dependency. Rich per-leaf filters bring enforcement close to municipal endpoints yet multiply configuration surfaces, audit scope, and drift risk across the fabric. Centralized choke-point steering reduces policy sprawl and eases change control, but it creates concentrated failure domains and potential bottlenecks for agency traffic that must hairpin. Distributed and centralized FBF do not always produce identical scale and failure domains; the topologies differ by design. It is false that only spines can host FBF so leaf placement is never valid—leaves are common attachment points in ERB civic fabrics. Centralized FBF also does not remove the need for filter matches because EVPN tags supposedly replace classification; if FBF is the chosen steering tool, classification remains required. Exam caveat: choose FBF placement from operational capacity and failure-domain goals, not from a myth that only one tier may host filters. Operational check: prototype both a per-leaf FBF set and a centralized steering design for a representative agency subnet, compare commit complexity, failover behavior, and troubleshooting time, then document the city’s chosen placement standard.