Dynamic spectrum access (DSA) scheme in Cognitive Radio (CR) can solve the current problem of scarce spectrum resource effectively, in which the unlicensed users (i.e. Second Users, SUs) can access the licensed spectrum in opportunistic ways without interference to the licensed users (i.e. Primary Users, PUs). However, SUs have to vacate the spectrum because of PUs coming, in this case the spectrum switch occurs, and it leads to the increasing of SUs' delay. In this paper, we proposed a Variable Service Rate (VSR) scheme with the switch buffer as to real-time traffic (such as VoIP, Video), in order to decrease the average switch delay of SUs and improve the other performance. Different from previous studies, the main characteristics of our studying of VSR in this paper as follows: 1) Our study is on the condition of real-time traffic and we establish three-dimension Markov model; 2) Using the internal optimization strategy, including switching buffer, optimizing buffer and variable service rate; 3) As to the real-time traffic, on the condition of meeting the Quality of Service(QoS) on dropping probability, the average switch delay is decreased as well as improving the other performance. By extensive simulation and numerical analysis, the performance of real-time traffic is improved greatly on the condition of ensuring its dropping probability. The result fully demonstrates the feasibility and effectiveness of the variable service rate scheme.

With the increasing popularity of cognitive radio networks (CRNs), providing data transmissions with quality of service (QoS) requirements in such networks becomes one of the most important topics. In this paper, the QoS provisioning in the CRNs with spectrum underlay has been studied. We propose the Joint Eviction and Admission Control (JEAC) for CRNs. Both real-time traffic (or, video traffic) and non-real-time traffic (or, data traffic) are considered. Because of real-time traffic is sensitive to average handoff delay more, our main goal is to ensure that its QoS requirement. For the non-real-time traffic, as far as possible to improve the throughput is our optimization goal.