A simple packet scheduling and buffer management scheme for scalable support of QoS in the Internet

We propose a scheme that provides scalable support of QoS in the Internet by aggregating flows in the backbone networks and using a simple and fast forwarding mechanism to handle aggregated flows in the backbone network routers. Four traffic classes are defined for such aggregated flows, one each for guaranteed, controlled-load, assured and best-effort services. In each router, each class is implemented as a separate queue, served using a priority-based discipline. Aggregated flows are shaped or marked before being inserted into the queues. RED-like buffer management is employed to control the buffer occupancy of conformant and non-conformant packets. A border router placed at the link between an RSVP-capable stub network and an aggregation region, translates the flows into these four classes by tagging the data packets. Simulations have shown that the scheme is able to provide rate and delay guarantees to individual flows even after aggregation.