Performance comparison between IntServ-based and DiffServ-based networks

It is generally believed that IntServ architecture is more suitable for strict QoS guarantees than DiffServ architecture because the former has per-flow-base traffic handling mechanisms but the latter does not. This general belief, however, has not been fully verified yet. In this paper, we theoretically and numerically compare IntServ and DiffServ architectures in terms of x-percentile end-to-end queueing delay as well as the maximum end-to-end queueing delay. We find that, in some network topology, DiffServ architecture makes the maximum queueing delay smaller than IntServ architecture does. We also find that in most cases x-percentile queueing delay in DiffServ-based networks is much smaller than that in IntServ-based networks