Fast Start Mechanism for Congestion Control of Networking

碩士 === 國防大學理工學院 === 資訊科學碩士班 === 98 === With the increasingly develepment of network technologies, the capacity of network bandwidth becomes larger and larger, but the high-speed network environment is limited by TCP transmission protocol and can’t effectively take advantage of the bandwidth. Most o...

Full description

Bibliographic Details
Main Authors: Wu,Feng-Nien, 吳豐年
Other Authors: Tang,Chi-Yi
Format: Others
Language:zh-TW
Published: 2010
Online Access:http://ndltd.ncl.edu.tw/handle/14608641781313981721
Description
Summary:碩士 === 國防大學理工學院 === 資訊科學碩士班 === 98 === With the increasingly develepment of network technologies, the capacity of network bandwidth becomes larger and larger, but the high-speed network environment is limited by TCP transmission protocol and can’t effectively take advantage of the bandwidth. Most of the TCP congestion algorithms are based on the RENO algorithm, which includes four parts: slow-start, congestion avoidance, fast retransmission, and fast recovery. The slow-start part is not very important for the local area network with few users and large bandwidth volumes. It has become a significant issue that how to maximize the usage efficiency of bandwidth during an initial short period for users. In an e-learning network environment, if a user can acquire enough bandwidth once the network links are initially connected, the delay can be effectively reduced. Therefore, for the slow-start part of the Reno algorithm, this paper proposes a fast-start concept to maximize the usage efficiency of bandwidth during an initial short period, so that the original long waiting time in the slow-start part can be improved. For the slow-start part and variation of congestion window while packet losing, this paper proposed the FSTP(Fast Start Transmission Protocol) to enhance the growth rate original congestion window in the slow-start phase and improve the waiting time to increase the effectiveness of network traffic. The suitable transmission rate by the receiving message of each cycle to adjust the congestion window size The experiment results showed that FSTP can not only fairly transmite but the traffics maintain a certain compatibility for the Reno algorithm. Additionally, FSTP significantly reduce the transmission efficiency and improved transmission reliability under an unstable network environment. In a large bandwidth and long-distance network environment, FSTP also increased the useage efficiency of network and voided the unnecessary waste of resources.