Performance Analysis of Fountain-Based Protocol in CSMA/CA Wireless Networks

The recent study [6] on fountain-based protocol (FBP) suggested that TCP nodes have incentive to switch to FBP approach in order to gain higher throughput with the trade-off of redundant packets. Moreover, it is reported that a network of FBP nodes only and sending at maximum speed will not degrade the overall throughput, as compared to the typical network of all TCP nodes. However, we find such claims are less appropriate in wireless networks where the packet collisions constitutes the majority of packet losses. The employed CSMA/CA scheme stipulates that all nodes to listen to the channel occupancy before initiating a transmission as to avoid the packet collisions in the networks. Since all packets have the same priority under CSMA/CA mechanism, the network bandwidth is fairly divided regardless of FBP and TCP. Based on the models of Wu et al. [10] and Altman et al. [3], we investigate the throughput of TCP and FBP under CSMA/CA wireless networks analytically and the analysis is verified again with ns-3. Our preliminary results show that nodes have weak incentive to switch to FBP in CSMA/CA networks when certain conditions are fulfilled.