Providing scalable QoS-sensitive services to applications with varying degrees of elasticity using aggregate-flow scheduling is a challenging problem. In a previous paper [Proc. IEEE/IFIP Int. Workshop Quality Service (2000) 211], we advanced a theoretical framework where an optimal differentiated services provisioning problem is formulated and solved to yield solutions for optimal per-hop control. In this paper, we extend our previous work by investigating performance evaluation and implementation issues associated with the induced optimal differentiated services architecture. We design a system that realizes the optimal per-hop control coupled with end-to-end adaptive QoS control, and implement a practical enhancement by introducing a scaling function which is applied to the TOS field label value in the IP header at each router. The scaling function allows the service provider to configure the per-hop control so as to export customized QoS separation-essential when shaping end-to-end absolute QoS over per-hop relative QoS-commensurate with the QoS profiles of the service provider's user base. We use simulation to study the structural and dynamical properties of differentiated services as affected by optimal aggregate-flow per-hop control.