A Network-Driven Multi-Access-Point Load-Balancing Algorithm for Large-Scale Public Hotspots

Part 1: Autonomic and Decentralized ManagementInternational audienceWireless networks are getting more and more popular and are a basic part of our life with the daily use of smartphones. Users expect high quality connectivity even in public spaces where a high number of clients connect to a limited spectrum on a geographically small area. Therefore, large-scale, high density wireless networks, like they are present at events, are getting more common, but provide a serious resource allocation challenge. Thousands of clients want to connect to a network consisting of multiple APs and a limited spectrum, while all of them should receive a decent connection quality, throughput and delay. Therefore, none of the APs should be overloaded, so that they can provide service for each connected client. The IEEE 802.11 standard stipulates that the client makes the decision to which AP to connect to. In high-density networks, the individual decision of the client can lead to an AP overload and oscillations in AP association as a client typically has limited information about the network performance and does not collaborate with other clients in taking its decision. This provides unwanted behaviour for load-balancing, as there is no control over the clients. Therefore, we present a method where the APs get control over the client and realise load balancing in such a network. The AP evaluates through a score if the client can connect and, if the client is connected, checks regularly if it is the best option for the client