This paper performs a comparative evaluation of two QoS architectures, RSVP Reservation Aggregation and Scalable ReservationBased QoS, aimed at providing QoS levels similar to the ones provided by the well-known RSVP/IntServ architecture, but scalable enough for use in high traffic core networks. The comparative analysis, based on each model's characteristics and on simulation results, shows that our Scalable Reservation-Based QoS architecture is able to support the same QoS guarantees provided by RSVP Reservation Aggregation, with significantly increased resource utilisation and a small penalty in signalling processing overhead.