Reliability and quality of service in opportunistic spectrum access

We first study the impact of the recovery time on the performance of the cognitive radio network. By classifying the failures into hard and soft, it is investigated how the availability, mean time to failure and mean time to repair are affected by the recovery time. It is observed that the time spent for recovery prevents the network from reaching the maximum availability. Therefore, to achieve a high mean time to hard failure and low mean time to repair, an available option is to increase the number of channels, so that with a high probability, a user who missed the channel can soon find a new channel. On the other side, an efficient recovery scheme is required to better take advantage of a large number of channels. Recovery improvement is thus indispensable.;To support traffic differentiation, we suggest a priority queueing approach. We extend the results of the general queueing model and discuss four different priority queueing disciplines ranging from a pure preemptive scheme to a pure non-preemptive scheme. New disciplines increase the flexibility and decision resolution and enable the CR node to more accurately control the interaction of different classes of traffic. The models are solved, so it can be analyzed how the reliability and quality of service parameters, such as delay and jitter, for a specific class of traffic are affected not only by the channel parameters, but also by the characteristics of other traffic classes.;The M/G/1 queueing model with interruptions is a foundation for performance analysis and an answer to the need of having closed-form analytical relations. We then extend the queueing model to more realistic scenarios, first with heterogeneous channels (heterogeneous service rate for different channels) and second with multiple users and a random medium access model. In the first part, the queue occupancy is modeled as a multi-row Markov chain where each row represents one of the possible service rates. In addition to numerically solving the Markov chain, two analytical approximations are provided. Although a Markov chain model necessitates assuming exponentially distributed channel availability periods, we further analyze and discuss the OSA queuing model for general distribution of service time and availability periods. The analytical and simulation results indicate that for usual system parameters, the queue average occupancy is similar for different distributions of service time and availability periods and that the exact memoryless Markov models can be used to accurately predict the heterogeneous OSA system traffic performance.;A multi-user scenario with discrete-time distributions is also investigated to have more insights on the impact of a baseline random medium access protocol and design of control channel on the performance of the cognitive radio networks. It enables us to study the interaction of recovery algorithms and the medium access protocol. It is observed that with an Aloha-type medium access, the performance can be better when instead of a switching recovery policy (i.e., vacating the channel in case of appearance of primary users), a waiting and buffering recovery policy is employed (i.e., the CR user waits for the primary user to vacate the channel) because in an Aloha-type medium access, the bottleneck is access to control channel and a switching policy necessitates more frequent control channel accesses.;To study the impact of recovery on higher communication layers, a queueing approach is chosen. Considering the recovery periods as a service interruption, a general M/G/1 queueing model with interruption is proposed. Different reliability and quality of service parameters can be found from this queueing model to investigate how channel parameters, such as availability and unavailability periods, and the recovery algorithm specifications, such as the recovery duration, affect packet loss, delay and jitter, and also the MTTF and MTTR for hard and soft failures.;While the main effort of this research work is to analyze the impact of spectrum handover recovery process, we also propose, a greedy and history-aware spectrum handover scheme to improve the time spent for spectrum handover (i.e., the recovery time). At the beginning of each timeslot and based on the state of the current channel, the scheme computes the optimal number of channels to be sensed in this restoration period and this number is dynamically updated after each channel sensing result. Intrinsic features of learning and history-awareness of CRs are used to create an optimal list of channels to be sensed based on the channels' background and historical information. Simulation results show that the history-aware sensing order improves the restoration mechanism by providing a shorter restoration time or a restored channel with a higher quality. (Abstract shortened by UMI.).