Performance modeling of multipath mobile data offloading in cellular/WiFi networks with bandwidth uncertainty

Mobile data offloading is one of the most effective approaches for improving the Quality of Service (QoS) of mobile users. Through this method, the mobile data traffic load over a cellular network is mitigated by offloading some of the traffic to an alternative path; usually a WiFi network. Most previous studies on WiFi offloading have assumed that the data traffic is offloaded to a path with constant bandwidth. However, in real-world applications, it is generally impossible to perform data transmission at a constant rate through WiFi over the entire offloading process, Hence, one of the most challenging issues in analyzing the performance of mobile data offloading is that of evaluating the effects of bandwidth uncertainty on the actual transmission rate. Accordingly, the present study proposes an analytical model for predicting the performance of multipath data offloading over WiFi network with bandwidth uncertainty. The theoretical results are validated by means of numerical simulations. The effectiveness of the proposed model is then further investigated by comparing the QoS metrics of the data offloading process with those obtained in a previous study on mobile data offloading under the assumption of a constant bandwidth. Overall, as the WiFi connection times are highly larger than the WiFi disconnection times, our model removes estimated errors in the baseline model by 8.7% and 6.7% in deadline miss ratio, by 23.2% and 14.9% in average offloading volume, when we consider two and three varying WiFi bandwidth respectively. The results show that the proposed method provides mobile operators with an effective framework for maintaining the QoS level of mobile data offloading in real-world networks characterized by WiFi bandwidth uncertainty.