In this dissertation, we address the problem of workload characterization in a wireless LAN (WLAN). Workload is generated by applications and users trying to carry out some of their functions. We attempt to capture such application- and user-level characteristics from the information gathered at the MAC level. Developing an understandable description of the workload requires making some abstractions at the application- and user-level. Our approach is to consider the workload in terms of ``sessions", where a session is an application- and user-level sequence of exchanges. We attempt to capture the session by considering an inactive duration in the activities between a wireless end-point and the network.

We consider workload to consist of a population of sessions for which a probability distribution function can be defined. Considering this distribution function to be a mixture distribution, we attempt to find the components by using non-parametric clustering technique. As the number of types of user level activities is not likely to be very large, we expect that we can associate a distinct activity with each such component. In this work, we identify such components and analyze the traffic and protocol characteristics of each component. Moreover, we empirically show that the identified workload components can effectively represent the actual WLAN workload and its daily variations.