Performance Enhancement for TCP over Dynamic Bandwidth Environment

• Main Authors: Chi-Len Chiou, 邱啟倫
• Other Authors: Neng-Chung Wang
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/81659764569993017310

碩士 === 朝陽科技大學 === 資訊工程系碩士班 === 94 === Transmission control protocol (TCP) is a reliable transport protocol in traditional networks which consist of wired links with low bit error rates (BER). In the traditional assumption for TCP, the packet loss was caused by the congestion at the relaying routers. The congestion at the relaying routers is the major reason of packet loss. Unfortunately, when TCP is used over wireless links, the performance of TCP would be significantly degraded. TCP is designed for many important applications such as web browsing and file transferring. The congestion control of TCP is based on an additive increase multiplicative decrease (AIMD) that adjusts the congestion window of TCP sender. The AIMD makes the sender to utilize the available bandwidth of current networks. TCP’s sliding window flow control mechanism often leads to burst packet traffic in the Internet. In the presence of network congestion, multiple packet losses would occur. It would degrade the performance of TCP and the TCP sender perceived the limit of networks. An excellent congestion control can intelligently set the congestion window (cwnd) and slow-start threshold (ssthresh) before a packet loss that affects the throughput of transmission significantly. In this thesis, we proposed a mechanism to dynamic adjust the slow-start threshold. The ssthresh estimation would set an appropriate ssthresh. The better ssthresh would improve the transmission performance of TCP distinctly. In the congestion avoidance state, we presented a mechanism to probe the available bandwidth. We would adjust the cwnd appropriately by consecutive observation round trip time (RTT) and reset the ssthresh after the fast retransmission or the timeout by the ssthresh estimation. Then, the TCP sender can enhance their performance by ssthresh estimation and consecutive observation round trip time. Our mechanism would define an efficient transmission rate and it could achieve higher utilization than other TCP versions. Simulation results show that our scheme can effectively improve the TCP performance. For example, when the average bottleneck bandwidth is close to 30% of the whole networks bandwidth, our mechanism improves the TCP performance by 10% at least.