Traffic Balancing: A Method for Exploiting System Capacity in Wireless Ad Hoc Networks

Carleton University, Ontario, Canada. November 2005.

Wireless Ad Hoc technology, which has received a rapidly increasing amount of attention over the last few years, provides a viable means of ubiquitous, untethered communication that could radically alter the way we work, learn, consume, and entertain. The routing protocols aim to set up connections and reestablish connections under a frequently changing topology. Current routing protocols generally search for the shortest path between sender and receiver, which usually results in fast response for route setup and a small number of hops. However, the shortest path algorithm has a high probability that traffic concentrates in the middle area of the network so that the system utilization is poor. Realizing the reality that the available system resources such as bandwidth in wireless Ad Hoc networks are limited and precious, research efforts have been undertaken to improve efficiency of resource utilization by means of traffic balancing. Because not all interferences are considered, these solutions fall short in collecting comprehensive traffic load information. Furthermore, some solutions generate a large amount of extra control packets that consume the available bandwidth.

 

The research in this thesis focuses on exploring the unused and wasted system capacity of wireless Ad Hoc networks, by providing comprehensive and accurate traffic load information to the routing protocols with minimum complexity. Proposed Traffic Balancing, a routing algorithm revised on the basis of reactive routing protocols, routes traffic load from congested areas to idle or lightly-loaded areas, such that network resources are allocated more efficiently. Knowing the number of observed collisions and past medium usage, every relay node provides the traffic load information into the route request in a simple way. After receiving route replies, a sender can choose the path with the least number of busy relay nodes. The simulation results illustrate that the Traffic Balancing approach is capable of decreasing the packet loss rate and average delay dramatically when some areas of the network start experiencing congestion with traditional on-demand routing protocols. Under certain scenarios, the improvement in system performance could exceed 50%. Furthermore, Traffic Balancing provides a solution to the problem of uneven traffic load that occurs at the access points of wireless mesh networks. The improvement is noticeable from the results of the simulation carried out on wireless mesh networks, even when the overall traffic load is light.

 
In this thesis, system capacity is first defined and investigated based on the interference range of a transmitting node and the size of network. The results show that the system utilization is far below the estimated capacity in the case where nodes are in movement. The major reason for the deficiency in system utilization is that movement and heavy traffic load cause collisions in congested areas, and the resulting long backoff periods after collisions waste a large amount of bandwidth.
It is also worth noting that traffic load tends to cluster, due to the routing and MAC protocols. Traffic Balancing is designed to provide accurate information about the traffic state along the paths, and to choose the path that has the least number of busy intermediate nodes. Numerical results indicate that the system performance is improved significantly when traffic load is deviated away from busy areas. With slight modifications, Adaptive Traffic Balancing is developed to change the verification rule of the traffic state at each node dynamically, according to the number of collisions detected by the node.


Furthermore, we investigate the problems that exist in wireless mesh networks, which have many similarities to wireless Ad Hoc networks. One severe problem is that performance is degraded by an uneven traffic load at the access points, whenever there is more than one access point in the network. After deploying Traffic Balancing, performance is improved and the system resources are used more efficiently.