Iscan: Efficient Wifi Scan For Mobile Device Based On Client And Network Behavior Learning
Xiaokang Sang,Qian Wu,Hewu Li
DOI: https://doi.org/10.1109/ISCC.2017.8024674
2017-01-01
Abstract:As the rapid growth of smart phones, dense and large-scale 802.11-based WLANs have become an important infrastructure to the Mobile Internet. Usually, mobile devices perform periodic scan to collect the information of surrounding APs for awareness of wireless environment. However, this kind of frequent scan severely degrades both the performance of network and client. In this paper, we first collect a real-world WLAN dataset which contains handoff information and related client state in two weeks from about 3500 mobile devices. Based on the analysis of this dataset, we observe that such kind of frequent scan for mobility is unnecessary for stationary client to pursue better performance, which will result in a large number of invalid scans and even the ping-pong. To address this issue, we propose an efficient WiFi scanning scheme called iScan, which triggers scan based on client and network behavior learning. The network behavior is represented by the change of network load, while the client behavior is detected by an accelerometer. In iScan scheme, when the client is moving, scans will be triggered if the RSSI is below the specific thresholds for different speeds. When the client keeps stationary, iScan adjusts the scanning intervals adaptively based on the law of network load change. We conduct experiments by both case study and simulation in campus WLAN. The case study shows that iScan can trigger scan more accurately without missing necessary handoffs when the client is moving. And the simulation shows that iScan reduces 95.8% of invalid scans while ensuring the stationary client discoveries the change of network load timely.