Fuzzy-based dynamic QoS parameters estimated by MMPP model for CBWFQ queuing

• Main Authors: Shi-Jie Yan, 顏士傑
• Other Authors: Fang-Yie Leu
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/34708370325326760038

碩士 === 東海大學 === 資訊工程與科學系 === 93 === As internet grow rapidly, network traffic become more and more heavy than usual. Network providers have provided more backbones and bandwidth for users, but user's requirements are always much more than supply. Therefore, network congestion becomes serious day by day. Quality of Service (QoS) seems one of the ways to solve this problem. Many QoS techniques have been so far raised. However, most of them just offer an approach or a platform, no exactly rule can follow up and no flexible system can manage varied network. This thesis proposes a network QoS system named Fuzzy-based Dynamic Resource Allocation System (FDRAS) which partitions network traffic into several channels and recommends suitable queueing strategies to network administrator so that different user traffic and heterogeneous data packets of different classes can be properly multiplexed. Class Based Weighted Fair Queueing (CBWFQ) is deployed as the congestion resolution mechanism. Markov Modulated Poisson Process (MMPP) is encompassed to model network traffic. FDRAS also invokes Expectation Maximization (EM) algorithm to estimate MMPP parameters and Fuzzy Theory to allocate the bandwidth reserved for a channel. Reserving bandwidth guarantees the least network traffic, while declaring threshold of queue limit defines the priority of a class. A packet with higher priority will be delivered with higher probability. Moreover, FDRAS can automatically tune its QoS parameters to adapt varied network traffic, thus decreasing managerial load and providing user a higher network service quality.