MASERATI: Mobile adaptive streaming based on environmental and contextual information

Abstract

Wireless/mobile video streaming has become increasingly popular, which makes wireless link bandwidth scarce. To provide streaming services to mobile users, it is crucial to adapt to the link condition and traffic fluctuation. We investigate which factors in natural environments and user contexts affect the available link bandwidth. To this end, we conduct a measurement study which contains 38 repeated trips along the same 5 km circular road in the campus of Seoul National University in April and May 2013. We measure the download throughput of video streaming from two different networks (3G and 4G LTE) with varying location, time, humidity, and speed. Our measurement results reveal that the humidity and location are the more important factors in the 3G network, while the speed, time, and location are the more important ones in the 4G LTE network to predict the available link bandwidth. We then propose an adaptive video streaming framework, MASERATI, where the information of environments and contexts is used to predict the available bandwidth. We demonstrate that MASERATI significantly improves the QoE of mobile streaming users in terms of the playout success rate, video quality, and stability, in comparison to DASH.