Abstract:
As the number of mobile devices accessing large-scale WLANs such as campus
and metropolitan area networks increases, the need for load balancing among the
cells becomes crucial. In addition, the network must also support some minimum
handoff tolerance defined by an application.
A number of load balancing techniques have been proposed in the literature that
focuses on formulating new load metrics rather than using Received Signal
Strength Indicator (RSSI) as the association metric. These schemes consider a
variety of factors such as number of STAs, enhanced RSSI, channel utilization,
queue length, bandwidth, and throughput to achieve balanced load. However,
some of these techniques require protocol modifications to both APs and STAs or
need special agents such as admission control server, extra software, and
switches. Others do not consider Quality of Service (QoS) requirements of
applications, which vary from one application to another, and thus do not satisfy
users requiring minimized handoff latency and real-time services. Moreover, most
techniques ignored the hidden node problem, which causes packet collisions and
thus the presence of such nodes can severely affect the performance of WLANs.
This dissertation proposes a new metric that provides load balance as well as
timely handoffs for WLANs by taking into account both direct and hidden node
collisions as well as the types of traffics in order to support QoS. Another novel
feature of the proposed method is the use of probe requests during the discovery
phase to monitor the states of the channels to determine the best Access Point
(AP) for association. Our simulation results show that the proposed method is
significantly better than relying only on signal strength in term of utilization,
end-to-end delay, collision rate, and packet loss.
