Research Article
Balancing Throughput and Fairness for TCP Flows in Multihop Ad-Hoc Networks
@INPROCEEDINGS{10.1109/WIOPT.2007.4480042, author={Yuedong Xu and Yue Wang and John C.S. Lui and Dah-Ming Chiu}, title={Balancing Throughput and Fairness for TCP Flows in Multihop Ad-Hoc Networks}, proceedings={5th International ICST Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks}, publisher={IEEE}, proceedings_a={WIOPT}, year={2008}, month={3}, keywords={Ad hoc networks Aggregates Degradation Distributed algorithms Feedback Loss measurement Performance analysis Proportional control Spread spectrum communication Throughput}, doi={10.1109/WIOPT.2007.4480042} }- Yuedong Xu
Yue Wang
John C.S. Lui
Dah-Ming Chiu
Year: 2008
Balancing Throughput and Fairness for TCP Flows in Multihop Ad-Hoc Networks
WIOPT
IEEE
DOI: 10.1109/WIOPT.2007.4480042
Abstract
Analyzing transport layer operation and enhancing its performance over multihop ad hoc networks have attracted a lot of attentions. Although the fundamental reasons for the performance degradation of TCP have been studied for many years, the tight coupling between transport layer and the wireless MAC layer is still not well understood. In this paper, we focus on the interactions between the hidden nodes and network congestion. By modeling the frame loss ratios of competing flows, a novel index is presented to measure congestion and fairness of these interfered links simultaneously. We formulate a practical optimization framework for TCP flows, and propose a distributed algorithm to improve the end-to-end throughput, and at the same time, provide per-flow fairness by exploiting cross-layer information. In the link layer, each node uses a proportional controller to determine the ECN marking probability for the purpose of notifying incipient congestion. Then the rate based TCP sender adjusts its sending rate according to the feedbacks from the link layer. Compared with standard TCP/802.11 as well as recent wireless TCP enhancements, our method substantially improves both long-term fairness and short-term fairness without sacrificing the aggregate end-to-end throughput. For some topologies with long hops, the throughputs of our proposed algorithm even outperform basic TCP/802.11 by over 100%.
