IEEE Transactions on Mobile Computing

Download PDF

Keywords

IEEE 802 11 Standard, Correlation, Prediction Algorithms, Throughput, Power Demand, Heuristic Algorithms, Mobile Communication, Link Aggregation, Wi Fi Aggregation, Multiple Radio Interfaces, Multipath TCP

Abstract

The concurrent use of multiple Wi-Fi radios in individual frequency channels is a solution readily available today to the increase of a mobile station’s communication capacity, but at the expense of occasional performance deterioration (when the heterogeneity of capacity between interfaces gets severe) and additional power consumption. This paper proposes a mobile-side solution for the concurrent use of multiple radios in a performance-aware and energy-efficient manner, with which a mobile station activates and deactivates radio interfaces dynamically according to traffic demands and a predicted capacity gain. To this end, the proposed solution is composed of multiple prediction algorithms and a control algorithm. Prediction when activating an additional radio interface is relatively difficult since no information of the disabled interface’s current status (and the corresponding frequency channel’s) is available at the time of prediction. Our experiments show that, despite different types and used channels, different radio interfaces have a strong correlation of received signal strengths and used PHY rates between them. Based on this observation, the proposed solution learns a correlation pattern between interfaces whenever multiple interfaces are active and makes prediction of the coverage, expected PHY rate and capacity impact of an inactive interface based on the learned correlation with a currently active interface. The design of the prediction algorithms are based on a simple or machine-learning technique (SVM). The control algorithm then keeps monitoring the utilization of active interfaces and, if any of them has utilization over a threshold, checks if each inactive interface is within coverage and a valid rate range based on an active interface’s received signal strength. Finally, an action of a configuration change (either activation, deactivation, or no change) selected based on the prediction of the resulting capacity is applied. Testbed experiments using COTS dual-band Wi-Fi interfaces demonstrate that the solution can enhance throughput by up to 29.6 percent (in a close distance to AP) and at most halve power consumption compared to legacy aggregation while the gain varies depending on the location and traffic conditions.
Like what you’re reading?
Already a member?Sign In
Member Price
$11
Non-Member Price
$21
Add to CartSign In
Get this article FREE with a new membership!

Related Articles