Section: New Results
Performance Evaluation in Wireless Nertworks
Participants : Fadila Khadar, Tahiry Razafindralambo, Marie-Emilie Voge.
Since the emergence of ubiquitous computing, evaluating wireless network performances has become one of the major economic issues. Among the existing performance indicators, the network capacity , defined as the maximal amount of flow carried by a topology during a fixed time period, is essential. Some cross-layer characteristics have to be taken into account in order to optimally allocate the common resources. In  , a comparative study is done between interference consequences in the two following models: (i) usual IEEE 802.11 MAC layer with acknowledgments at each hop, and (ii) block acknowledgments reported at the transport layer that can be included in the IEEE 802.16 standard. Cross-layer properties are modeled in a linear programming formulation that is solved using the column generation process. We quantify the gain in capacity induced by the move of the MAC acknowledgments into the transport layer, and show the better load distribution obtained in the network with the second model.
The performance of ad hoc networks based on IEEE 802.11 DCF degrade when congestion increases. The issues concern efficiency and fairness. Many solutions can be found at the MAC layer in the literature, but very few solutions improve fairness and efficiency at the same time. In this In  , we design a new backoff solution, called SBA. SBA uses only local information and two contention window sizes. By simulations, we compare SBA with IEEE 802.11 and several alternatives to 802.11 in ad hoc networks. The results presented in this paper show that SBA is fairer than 802.11. Of course, the fairness induced by SBA reduces the global throughput (compared to 802.11), but this decrease is often lower than the one achieved by some other fair MAC protocols. Given the first obtained results, we think that SBA is a good candidate in terms of trade-off between fairness, simplicity and efficiency.
In  , we consider Wireless Sensor Networks (WSN) applications in which sensors have to send data to a unique sink in a multi-hop fashion. The routing problem in WSN has been the subject of intense studies. One important difference between wired and wireless networks is the use of location for routing purposes. Position awareness improves the efficiency and scalability of routing protocols as it helps reducing the number of messages used for route discovery. Gradient routing protocol uses virtual coordinates, i.e., a coordinate system set up for routing purposes only. They create a one-dimensional virtual coordinate system where the position of a node corresponds to its hop distance to the sink. This information is then used to efficiently route packets to the sink in a multi-hop fashion. Many gradient routing protocols exist, they mainly differ in their performances (delay, delivery ratio, etc.). In this paper, we propose an extensive performance evaluation study of some gradient routing protocols and show that the choice of the algorithm must be based on what is expected from the network (reliability, energy consumption, ...). .