| บทคัดย่อ(English) |
One of the most important tasks for a policer in ATM networks is tomeet Quality of Service (QoS) requirements for different classes ofservice. The conventional token leaky bucket, which is designed forpolicing single traffic, cannot perform well in dealing with differenttraffic types and guaranteeing their QoS. As a result, the priority leakybucket has been proposed to cope with this problem by providing prioritylevel for each service class according to its requirement. However, thescheme still has a great disadvantage that the QoS of low priority class isdegraded as a tradeoff for better QoS of high priority class. So, thisthesis proposes two new priority leaky bucket mechanisms in order toimprove the QoS of both traffic classes. The first new mechanism adds anextra token pool with the conventional one that can improve QoS of bothpriority classes and resource utilization. The second one applies a fuzzylogic system with the first mechanism to adjust the token generation ratedynamically, according to traffic source characteristics. Moreover, thisthesis also studies the selectivity and responsiveness of the prioritymechanism when traffic is overloaded. These important properties of themechanism as a policer have never been investigated in the past priorityscheme literature. The simulation results show that when the traffic source respects theservice contract, the first and the second proposed mechanisms provide 30 %and 90 % better QoS than the conventional one, respectively. When onetraffic class violates the contract, the proposed fuzzy logic mechanismshows that it can control the effect of the excessive cells and stillmaintain the QoS of the other traffic class in the finest level whencompared to the other mechanisms. The proposed fuzzy logic mechanism alsohas a fast responsiveness, that is, it firstly detects cell violations 26%faster than the other mechanisms. |
