Wireless technologies have come a long way, from the radio technologies that allow data bits to be exchanged between two physically-disconnected devices, to the multiple access technologies that allow a group of devices to share a common wireless communication channel, and to the routing technologies that allow out-of-range devices to communicate via multihop wireless paths. During the recent years we have witnessed the huge commercial success of WLAN, which bloomed into a multi-billion dollar market. Multihop wireless networks, including wireless sensor networks, wireless mesh networks, and mobile ad-hoc networks, are expected to lead in the next wave of deployment.
However, the current technologies have limitations that serve as barriers to the widespread applicability of multihop wireless networks. Dr. Shigang Chen, an assistant professor with the Department of Computer and Information Science and Engineering, recently showed that some basic technologies such as MAC-layer fairness that contribute to the WLAN success cause end-to-end performance problems in multihop networks. More importantly, from the user's perspective, not only must the networks provide a robust and efficient communication infrastructure, but also they should provide means to support the diverse requirements at the application level, particularly, the ability to differentiate various types of end-to-end traffic flows and ensure quality of service, which is an open problem for multihop wireless networks.
Winning an NSF CAREER award in 2007, Dr. Chen and his research group have been working on distributed algorithms and network protocols to solve several fundamental problems in multihop wireless networks, including end-to-end weighted bandwidth allocation, bandwidth assurance, and performance/overhead tradeoff in traffic differentiation. The study covers a variety of network conditions, including single-commodity or multi-commodity flows, single-path or multi-path routing, and static or highly-dynamic wireless networks. Their research focuses on two hop-by-hop, routing-independent, and light-weight approaches based on aggregate state and packet labels, respectively. Without maintaining any per-flow state, these approaches are able to implement traffic differentiation under aggregate or weighted maxmin models and have great flexibility in self-adaptation based on changing network/traffic conditions.
Traffic differentiation is a difficult problem in wireless networks because of resource constraints, dynamic link capacities, multipath routing, dynamic topology/flows, and lossy communication channels, which make the solutions in wired networks not applicable. But the reward of this research will be enormous. It has many important applications in practice. First, let's consider a sensor network that collects data from a field. When an important event (such as fire outbreak, earthquake, or enemy movement) triggers a surge of data traffic from hundreds or even thousands of sensors to a small number of base stations, the network can be severely congested, especially near the base stations. Traffic differentiation allows sensors at the important locations to report data at higher rates and sensors at other locations to report at reduced rates. With this capability, the user can obtain an adjustable view of the monitored field with different resolutions at different locations. Second, traffic differentiation can promote the adoption of wireless mesh networks by home users. A user who carries a lot of others' traffic would demand (and deserve) a larger share of network bandwidth than a perimeter user who does not carry other's traffic. Multimedia applications with bandwidth requirements should be given priority over applications with elastic bandwidth demands as long as the benefited users give up certain bandwidth at later times to keep their long-term averages. Third, consider a mobile ad-hoc network deployed in a battlefield. Data flows between commanders and their units are more important than flows between soldiers; reports about approaching enemy tanks are more important than reports about foot soldiers. In this environment, traffic differentiation will ensure the delivery of critical data amidst an overwhelming amount of less important information.
Wireless sensor networks, wireless mesh networks, and mobile ad-hoc networks will provide a pervasive communication infrastructure for the modern societies and dramatically change the way people interact with the cyberspace and the physical environment. The proliferation of wireless networks in houses, workplaces, shops, cars, and highways will improve the quality, safety, and productivity of our lives. Multihop wireless networking is also a major emerging market. Dr. Chen's research on traffic differentiation allows these types of networks to meet diverse application requirements, which will promote their entrance into the marketplace.
Traffic differentiation in multihop wireless networks is drastically different from its counterpart in the wired networks forming today's Internet; yet the literature devoted to this subject is very limited. Dr. Chen's research team is working to fill this gap through a systematic study of the problem and solution space. Its success will stimulate further research in this and related areas in computer networks.
