Active Queue Management with Weights to Provide Differentiated Services

• Main Authors: Zhi-Yi Chen, 陳志毅
• Other Authors: Lain-Chyr Hwang
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/15940348252466896781

碩士 === 義守大學 === 電機工程學系碩士班 === 95 === Nowadays, the multimedia networks are common. To satisfy the different requirements in the multimedia networks, the QoS control mechanisms are necessary. Among these, the active queue management (AQM) plays an important role. Therefore, this thesis modifies the algorithm “GREEN (Global Random Early Estimation for Nipping).” GREEN enhances the concept of “Early” to “Early Estimation” by early estimating the network status. The network status obtained by estimation is earlier than by detection. In this way, the decision in the AQM can be made more quickly. After modification, the algorithm iss more stable than original GREEN and still keeps the original good features. This thesis also proposes a novel AQM algorithm named PASS (drop with Probability And StatisticS). In general, AQM should drop packet with random. We classify the random manner into two types: probability and statistics. For TCP flows, PASS drops the packets with probability, and for UDP flows PASS drops packets statistics. This method should works even for the networks without DiffServ. PASS uses the “match drop” like CHOKe and CHOKeW. But the difference is that the match drop of PASS just uses for UDP flows and there are no comparisons of flow id. In this way, the throughput of unresponsive flows is suppressed to protect TCP flows like CHOKe and CHOKeW. Furthermore, the overall algorithm is more efficiency.