A branch has a 50-Mbps internet circuit carrying voice and bulk replication where excess replication bursts cause drops for voice during congestion. Replication must be capped at 20 Mbps with excess dropped, while total egress must be smoothed to the circuit rate without discarding conforming voice. Which shaping and policing treatment meets this congestion requirement?

Policing enforces a hard rate by dropping or remarking excess packets without buffering, which suits capping replication at 20 Mbps, while shaping buffers excess into queues to smooth output to the 50-Mbps circuit rate, which preserves conforming voice during bursts. Reversing them fails by concept because shaping replication would buffer rather than cap bulk traffic and policing total egress would indiscriminately drop voice plus data exceeding 50 Mbps instead of smoothing. Policing both to 50 Mbps fails because edge policing without shaping causes tail drops for bursty egress and provides no per-class 20-Mbps cap for replication. Shaping both without dropping fails because shaping alone never discards, so replication could still occupy buffers and delay voice rather than being capped. Exam caveat: remember policing drops and shaping delays, and practical check is show policy qos counters for policer drops and shaper queue depth on the WAN Edge.