A Load Balancing Method In Enterprise Wireless LANs Using User Social Relations

With the rise of smart phones, user demand for mobile Internetaccessing has been increasing. In the face of growing number of accessedusers for the wireless local-area networks (WLANS), AP load balancingproblem become more important. Well balanced workload amongwireless access points in WLANS can improve the user experience foraccessing the Internet.Most load balancing research work in WLANS focuses on theoptimization of AP operations assuming that the arrivals and departures ofuser are independent. However, through the analysis of AP usage log on areal WLAN trace, we found in Enterprise WLANS, the relations betweenuses are much closer and this assumption does not hold. In facts, due tothe social relation between users, they tend to arrive or leave in unison,which would affect the load balance among APs.We analyze the distribution of social relation networks in a Chinesefamous university and propose a novel AP allocation method. We prove the load balancing problem in WLANS is NP hard. and propose an onlinegreedy scheme In this scheme, user with intense social relationships areassigned to different APs so that jointly leaving of those users would haveminor impact on the load balance of APs. Its computational costis O (n2logn m), and approximation factor ismm1, here m is the numberof APs and n is the number of users.By the real data driven experiment, we compare the performancebetween social relations based method and least load first method.Experimental result shows that, the performance for our method can beimproved by up64.7%, and the average balancing index can be improvedup to24.2%. And the performance for our method is much more stable.