Data transport design in challenged wireless networks
Mário Gerla, Guang Yang · 2006
Growing popularity of wireless networks has spurred increasing interest in data transport design and evaluation for these new scenarios. Wireless networks have their unique characteristics, which make legacy protocols originally designed for wired networks unsuitable. For example, most wireless links are error-prone. Node mobility is another issue commonly seen in wireless networks. These factors make wireless networks error-prone with frequent random packet losses. Legacy TCP assumes that every packet loss is caused by network congestion. It reacts by cutting its window to reduce the sending rate and is inefficient in the presence of random errors. To improve its performance, we propose a solution called TCP Bulk Repeat (BR), equipped with three sender-side modifications, namely the Bulk Retransmit, Fixed Retransmit Timeout and Intelligent Window Adjustment, together with a Loss Discrimination Algorithm (LDA). TCP BR can recover multiple losses faster, avoid unnecessary retransmit backoffs, and keep its window at a reasonable size. It can improve the TCP throughput by an order of magnitude in very lossy conditions. Streaming is another important application in wireless networks. Popular multimedia transport schemes such as the TCP Friendly Rate Control (TFRC) aims at delivering a throughput equivalent to TCP NewReno, and is as inefficient as NewReno in the presence of random errors. To achieve efficiency and smoothness while being TCP friendly, we propose the Video Transport Protocol (VTP) with a unique rate control mechanism. VTP relies on two key techniques: Eligible Rate Estimation (ERE) and Loss Discrimination. VTP is as friendly as TFRC. In addition, it maintains smooth and efficient rate control in both error-free and error-prone scenarios, and is suitable for real-time streaming across wireless networks. Beyond the conventional end-to-end data delivery discussed above, some special and challenged wireless scenarios require new transport models. We look at the connectivity disruptions in Mobile Ad-hoc Networks (MANETs) as a Delay-Tolerant Networking (DTN) problem and propose an Ad-hoc Storage Overlay System (ASOS) to cope with such disruptions. ASOS is an overlay providing reliable transport and redundant, distributed data storage. It is specifically designed for MANETs and complements the approaches that actively bridge disconnected areas with additional nodes.