الفهرس 
introductionOnly 14 pages are availabe for public view
 |  | from | 60 |  |  |
| | | from | 60 | | |
Abstract
Transmission Control Protocol (TCP) is the dominant transport protocol in the Internet. It supports most of the popular Internet applications, such as file transfer, e-mail and etc. Recently, Internet applications have evolved from standard document-retrieval functionality to advanced multimedia services. The rapid growth of the Internet and the increasing of the traffic demand led to a serious problem called congestion collapse.
Over years, continuous efforts have been done to avoid this problem, and many congestion control mechanisms are developed and refined by researchers aiming to overcome congestion. These mechanisms include Tahoe, Reno, NewReno and SACK. Although several variants were developed and refined, most of them employ Additive Increase Multiplicative Decrease (AIMD) strategy for adjusting the congestion window size (cwnd). In other words, they use the AIMD to change the rate at which packets are injected into the network. This strategy is inefficient in terms of utilization of link capacity and unfair in throughput. This is because, the AIMD strategy blindly updates the congestion window size statically by a fixed value regardless the network status. That is, the mechanism increases the congestion window by one segment as long as no congestion is detected, and decreases the congestion window to half of its value, as long as congestion is detected.
This thesis first presents the congestion problem and investigates the congestion control algorithms. Then, it presents a new strategy based on fuzzy logic to be used with End-to-End congestion control protocols instead of the AIMD strategy so as to improve their performance. The basic idea is to adapt the congestion window size dynamically based on the available network capacity. The proposed strategy is evaluated by using the network simulator NS2 and compared with the most widespread congestion control protocols; TCP Tahoe, Reno, New Reno and Sack. The simulation results show that the proposed mechanism improves the network performance against throughput, packet drop, packet delay and fairness.