Analysis of a buffered cognitive wireless network with dynamic spectrum assignment

In a cognitive wireless network, the sudden arrival of a primary user (PU) can force one or more secondary users (SU) to terminate their ongoing communication. Buffers can be utilized to prevent them from dropping, but their effectiveness depends on the tolerance of the SUs to the buffer waiting time. In this paper, we propose to dynamically assign service rates to the SUs to complement the gain offered by the network buffers in reducing the dropping probability of the SUs. Our analysis is based on a two-dimensional Markov chain with four state variables. Performance metrics for the SUs such as dropping, blocking and completion probabilities, and the average time spent in buffer are derived. Our analytical results demonstrate that depending on the network condition, the proposed dynamic spectrum assignment scheme is alone capable of reducing the dropping probability substantially. Since the sum of call completion, blocking and dropping probabilities is 1, the reduction of dropping probability of the SUs is achieved by increasing the sum of call blocking and completion probabilities, which not necessarily implies higher blocking or reduced completion probabilities.