标题: | 讯务塑型器-使用参数控制的深入研究与差异性服务网路新的量测机制 Further Investigation of TS-UPC and New Metering Algorithms for DiffServ |
作者: | 黄庆喜 Ching-Shi Huang 张仲儒 Prof. Chung-Ju Chang 电信工程研究所 |
关键字: | 使用参数控制;量测器;标记器;单速率三色标记器;双速率三色标记器;时间滑动视窗三色标记器;usage parameter control;meter;marker;single rate three color marker;two rate three color marker;time sliding window three color marker |
公开日期: | 2000 |
摘要: | 在这篇论文中,我们更进一步探讨在非同步传输模式网路更多变的环境下乏晰讯务塑型器(TS)-使用参数控制(UPC) 与类神轻乏晰讯务塑型器(TS)-使用参数控制(UPC) 的强健性。乏晰TS-UPC是由传统漏水桶法与乏晰水增额控制器(FIC) 所构成的,FIC利用此连线的长期平均速率和短期平均速率来调整水增额的量。类神经乏晰TS-UPC是由传统漏水桶法与类神经乏晰水增额控制器(NFIC)所构成的,同样的NFIC也是采用此连线的长期平均速率和短期平均速率来决定水增额的量,而且它有一个额外的强制讯号(reinforcement signal)使的它有线上(on-line learning)学习的能力,可以更精确的反应讯务的情形。模拟结果显示类神经乏晰TS-UPC与乏晰TS-UPC效能表现都优于传统TS-UPC,而类神经乏晰TS-UPC又优于乏晰TS-UPC。 近来有人提出差异性服务(Differentiated Service)网路架构以提供不同程度的服务品质保证。在差异性服务网路架构下,位于边缘路由器(edge router)的讯务调节器(traffic conditioner)负责监测进来的汇集讯务,并且根据监测结果来给予讯务相对应的处理。量测器(meter)是讯务调节器重要的一部分,它是用来量测进来的汇集讯务,以确保汇集讯务不会超过讯务合约。量测器和使用参数控制器扮演相同的角色,但不同的是量测器是对汇集讯务作监控的。在这篇论文中,我们提出增强型的双速率三色标记器(TRTCM)与增强型的时间滑动视窗三色标记器(TSWTCM),它们均可保护高优先权的封包不受低优先权的封包的影响,以及当有足够的资源时允许提高封包的优先权藉以改善频宽的使用效率。增强型的TRTCM是由两个图腾桶法(token bucket)所构成的,而对于封包作标记则是由图腾使用的情形以及进来的封包标记来决定。增强型的TSWTCM是由一个速率量测器(rate estimator)与一个标记器(marker)所构成的,速率量测器负责量测汇集讯务中不同优先权封包的速率,标记器根据量测的速率和最高速率(PTR)与协定速率(CTR)作比较后再决定对封包的标记。在模拟结果中,增强型的TRTCM效能表现和增强型的TSWTCM差不多,但是增强型的TRTCM和增强型的TSWTCM效能表现明显优于TRTCM。 In this thesis, we further investigate the robustness of the fuzzy TS-UPC and the neural fuzzy TS-UPC under more variant environment in ATM networks. The fuzzy TS-UPC is composed of the conventional leaky bucket and the fuzzy increment controller (FIC). FIC is used to compute the appropriate increment value according to the long-term mean rate and the short-term mean rate. The neural fuzzy TS-UPC is composed of the conventional leaky bucket and the neural fuzzy increment controller (NFIC). NFIC is used to intelligently compute the increment value according to the long-term mean rate and the short-term mean rate. In NFIC, the reinforcement learning can learn on line from a reinforcement signal (the difference between desired loss ratio and measured loss ratio), so the increment value can reflect accurate traffic condition. Under the more variant environment, simulation results show that the neural fuzzy TS-UPC and the fuzzy TS-UPC both perform better than the conventional TS-UPC, and the neural fuzzy TS-UPC outperforms than other TS-UPCs. Recently the Differentiated Service (DiffServ) model has been proposed to provide different degrees of QoS guarantee. The traffic conditioner located in the edge router monitors the aggregate traffic of incoming packets and takes corresponding actions according to the monitoring results. One important component of the traffic conditioner is the meter. It is used to monitor the aggregate traffic of incoming packets to ensure that the aggregate traffic would not exceed negotiated traffic profile. It plays the same role as UPC, but the difference between the meter and UPC is that the meter works at the aggregate traffic. We propose the enhanced TRTCM (Two Rate Three Color Marker) and the enhanced TSWTCM (Time Sliding Window Three Color Marker) that not only protect high priority packets from affecting by low priority packets but also improve the utilization by allowing promotion of packets when there is available bandwidth. The enhanced TRTCM consists of two token buckets and decisions on marking of packets depend on the token usage of the two token buckets and the colors of incoming packets. The enhanced TSWTCM consists of a rate estimator and a marker. The rate estimator measures the incoming rates of different priority packets within a fixed window size, and the marker determines the marking of packets based on the measured rates as compared to the PTR (Peak Target Rate) and the CTR (Committed Target Rate). Simulation results show that both the enhanced TRTCM and the enhanced TSWTCM outperform the TRTCM. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT890435018 http://hdl.handle.net/11536/67298 |
显示于类别: | Thesis |