We propose a dynamic QoS, or D-QoS, model where QoS settings can be automatically reconfigured based upon requests from authorized users. Different levels of privilege can be assigned to users enabling higher privileged users to interrupt the network flows belonging to those of lower privileged levels. To request for a special QoS treatment, a user can issue an active packet to interrupt any active node along its flow path which is D-QoS enabled. The request for a specific interruption level is approved by a D-QoS enabled node which allows for multi-level interruptions to be handled. After an interrupting flow has completed transmitting all its packets, D-QoS enabled node can resume its services for those pending flows which are of lower privilege levels. In this paper, we describe the overall concept of D-QoS and demonstrate how it can be implemented by a small prototype. Using simulation, we show that the proposed system can provide assurance for privileged flows with an improved network utilization where bandwidth is shared among the flows according to the levels of privilege. D-QoS should be deployed on those bottleneck hops with limited bandwidth on the edge network to ensure the best service is given to privileged users.