A lot of research work has recently focused on the exploitation of the DiffServ framework towards building reliable networking services that provide deterministic quality guarantees. In this work we are attempting to devise a DiffServ-based service model for relative differentiation of traffic in a network serving aggregated traffic. The service model proposed aims at providing a minimum capacity during congestion and a bounded average end-to-end delay to profile-conforming traffic, obtaining thus a relatively better performance when compared to the best-effort service. We provide the framework on which the proposed service model is based and we present results from its deployment in a simulation environment.