Characterizing sociality for user-friendly steady load balancing in enterprise WLANs

Traffic load is often unevenly distributed among the access points in enterprise WLANs. Such load imbalance results in sub-optimal network throughput, unfair bandwidth allocation among users, and unsatisfactory user quality of experience. We have collected real traces from over 12,000 WiFi users at Shanghai Jiao Tong University. Through intensive data analysis, we find that the social behavior of users (e.g., leaving together) may cause a significant AP load imbalance problem. We also observe from the traces that users with similar application usage have the potential to leave together. Inspired by those observations, we propose a socialaware AP selection scheme (S³), which can actively learn the sociality information among users trained with their history application profiles and elegantly assign users to different APs based on the obtained knowledge. Trace-driven simulation results show that S³ is feasible and can achieve better balancing performance when compared to state-of-the-art balance algorithms.