A network engineer reviews the QoS policy on an enterprise WAN aggregation router. During periods of congestion, users complain that many TCP sessions slow down simultaneously and then recover together, causing throughput to oscillate on the link. The engineer wants to configure a congestion-avoidance mechanism that selectively drops packets before the queue is completely full to prevent this synchronized behavior. Which QoS mechanism should the engineer configure to address this problem?