Comprehensive performance model of differentiated service with token bucket marker

Differentiated services (DiffServ) networks use two type of routers: edge routers mark the packet headers according to their service profile, while the core routers implement queue management to provide service differentiation between packets according to information in their packet headers. Most previous studies on the performance of DiffServ have considered either the effect of the edge marking scheme or the effect of core queue management on the throughput of TCP applications. To study the combined effect of marking at the edge router and queue management at the core router on the performance of DiffServ, the authors propose a simple model that takes into consideration the committed information rate and maximum allowed burst size of the marker, in addition to the number of flows in an aggregate, round trip time, and the probability of a packet dropping at the core network. Observations from the model are presented, and simulations were performed to validate the accuracy of the model. The model shows that the performance of DiffServ is largely flow-based rather than desired-aggregate-based, and the DiffServ network suffers from unfairness issues.