The tremendous success of 802.11b wireless networks ensures such wireless network deployments will become ubiquitous in the near future. An ad hoc wireless network based on ubiquitous and low cost 802.11b interfaces would be an ideal candidate for disaster relief applications during times when traditional networks fail to provide adequate network services. However, 802.11b was not designed with providing QoS in mind, and it is not trivial to provide QoS guarantees over such networks. To bridge the gap, we investigate the problem of providing soft QoS guarantees in a small sized 802.11b ad hoc wireless network.
One of the major challenges in providing QoS is in deciding routes for application traffic. At the heart of this routing problem is finding the right model for describing the shared nature of the wireless medium. We used a simplified model for the delay and bandwidth constraints of a sharing medium based on the use of a cell. We proved, through an implementation, that this simple model captures the important charateristics of the shared medium constraints specific to wireless routing networks. The simplicity of the model allows us to use an integer linear program to find the optimal routes. Furthermore, the routes chosen by using the simple model are verified by a faster than real time simulation that can use more sophisticated interference models that can provide more accurate results but are much harder to analyze. In order to cope with the inevitable differences between the simple model and the more accurate MAC layer simulation model, we invented a control loop that can search and adapt feasible QoS routes, starting from those calculated from a simple model by incorporating a simulator in the loop. In addition, to compensate for the lack of MAC layer reservation capabilities, we introduced a simple distributed scheduling algorithm to rate limit the best effort traffic to protect the QoS requested by high priority applications. Finally, we completed a proof of concept implementation of all of the above features to show that our architecture is both self-contained and sufficient to provide QoS in an ad hoc wireless network using only off-the-shelf 802.11b network equipment.