Since the recent explosive expansion of internet users and applications has made quality of service(QoS) requirements for various applications such as voice and video over IP. The IETF proposes the differentiated services(DiffServ) for satisfying these QoS requirements. The DiffServ classifies incoming packets into three service classes which are expected forwarding(EF), assured forwarding(AF) and default(DE). It offers the proper QoS to each class according to per-hop behaviors(PHBs). And a scheduler is needed for scheduling these multiple classes.
There are two traditional scheduling methods known as PQ(Priority Queue) and WRR(Weighted Round Robin). However, these two scheduling methods have some drawbacks. Because PQ services the higher priority class above all the lower classes, if the higher priority class is always full then the lower classes can not be serviced. In WRR, each class uses the link capacity by the assigned weight so that each class is fairly serviced. When the EF class traffic is heavy, however, WRR can not satisfy the QoS requirement of the EF class traffic.
To overcome these drawbacks of PQ and WRR, propose a new scheduling method named the modified WRR mode 2(MWRRM2) based on two basic ideas. First, when there exist surplus link capacities, the capacities are used by EF above the other classes. Second, when the EF input traffic rate is increased then the EF class borrows some capacities from DE and returns the capacities to DE which are remained after use.
For the simulation of the proposed scheduling method in this work, we use the Poisson process and the self-similar process for source traffic modeling. The simulation show that MWRRM2 maintains the required QoS of the AF class when the EF traffic rate is increased and improves QoS of the EF class, too, at the price of the performance of the DE class traffic.
