A Study On Differentiated Access Points Selection In WLAN

Today, IEEE 802.11 WLAN (Wireless Local Area Network) is a popular connectivity method, as its ease of deployment, low cost and flexibility. WLAN is commonly deployed in airports, hospitals, cafes, libraries and so on. With the WLANs’deployment of high-density and large-scale, it’s common for users facing multiple available APs to associate with. Currently AP selection in 802.11 protocols is based on the RSSI (Received Signal Strength Indication). However, this simple strategy causes low network performance and load imbalance.Thus, AP selection optimization becomes a key issue since it will dictate the user’s actual performance.The existing AP selection policies proposed recently have the following drawbacks:first of all, they provide the same selection metric for all users neglecting the differences of their QoS requirements; secondly, the majority approaches base on single objective maximization while may not work in all circumstances and all kinds of service; lastly, some selection metrics computation is so complex and the process is repeated in every users, leading increased association time cost.Focusing on the problems mentioned above, this paper proposes differentiated AP selection strategies in WLAN. Firstly, it reseaches the QoS requirements of applications, then it classifies users according to their application’s different requirements on bandwidth and delay. Based on that, firstly a differentiated AP selection strategy using the Analytic Hierarchy Process (AHP) is proposed. It uses the AHP to derive multiple objectives maximized AP selection metric for users of each type respectively; at the same time, it implements (periodic and aperiodic) dynamic selection due to the network state changes and user application changes. Secondly, an improved differentiated AP selection approach using collaborative filtering recommendation is proposed, In a period of time, the first user of each type successfully associated with an AP, becomes the pioneer user of its type. When a new user joins the network, it recommends AP for the new comer depending on the association information of the pioneer user of same type if the former is judged as a near neighbor of the latter. Thus with a success recommendation the new user can associate with the recommendated AP immediately exempting from channel scanning and selection metric computation. At the same time, load pre-balancing module periodically exclude the AP considered over-loaded from the recommendation list, while the pioneer user adaptively updating module periodically replaces the AP considered to be low in recommendation ability.The simulation experiments demonstrate that the strategies:can provide differentiated QoS assurance to users; can improve the network overall throughput and highly redistribute network load; can degrade association time cost.