Average Waiting Time of Packets with Different Priorities in Cognitive Radio Networks Hung Tran, Trung Q. Duong, and Hans-J¨ urgen Zepernick Blekinge Institute of Technology, SE-372 25, Ronneby, Sweden E-mail: {htr,dqt, hjz}@bth.se Abstract—In this paper, we investigate the average waiting time of packets with different priorities in cognitive radio networks (CRN) using a preemptive priority queuing system. Specifically, we consider two scenarios for CRN, the first with the secondary user (SU) sensing at the beginning of each time slot and the other with the SU having continues sensing ability. Our analysis shows that the average waiting time of packets for the SU does not only depend on the size of packets and arrival rate of the SU traffic but also depends on the arrival rate and size of packets from primary users (PU). Moreover, the results show that an SU with continuous sensing ability can utilize spectrum better than sensing at the beginning of each time slot. I. I NTRODUCTION Being a limited resource, radio spectrum becomes exhausted with the growing of new mobile applications and the com- pelling need for mobile Internet access. However, the mea- surement at intensively crowded areas shows that only small parts of licensed spectrum are used at certain time [1]. In the effort to improve the utilization of limited spectrum resource, cognitive radio network (CRN), first proposed by Mitola [2], has gained great attention in the research community. In a CRN, there are two types of users, namely, primary users (PU) and secondary users (SU) or cognitive users. In the CRN concept, the transmission channel is licensed to the PU while the SU opportunistically accesses the channel resources when it is not used by any PU. Accordingly, an SU is required to detect exactly a vacant channel for transmission and vacate it when a PU wishes to use the channel. This transmission strategy is called opportunistic spectrum access [3], [4]. From the viewpoint of a PU, an SU should operate transparently to the PU but should not cause any interference to the PU. On the other hand, the SU operates in an unstable environment meaning that it’s transmission can be corrupted any time by the random access channel of any PU. As a consequence, the SU has to face some delay in the delivery of packets which in turn translates into waiting time. In view of the above, studies on average waiting time of packets in CRN have been given large attention. Specifically, the average waiting time of packets for SU is studied in [5], [6] under various preemptive priority queueing models such as M/D/1 (Poisson arrival process, deterministic distribution of service time, single server) and M/G/1 (Poisson arrival process, general distribution of service time, single server). However, these works do neither examine the role of the size of packets nor the impact of different priorities of packets on the average waiting time. In [7], a virtual queueing model is reported that examines the impact of packets with different sizes and priorities on average waiting time but for frequency division multiple access (FDMA) while the derived formula describe approximations. In this paper, we derive the average waiting time of packets with different priorities for PU and SU in the context of time division multiple access (TDMA) by employing a preemptive priority M/G/1 model. Furthermore, we show that the average waiting time of packets for an SU not only depends on the arrival rate and service rate of packets but also depends on the size of the packets in both SU and PU. To the best of our knowledge, this is the first paper analyzing this problem. The remainder of this paper is organized as follows. The system model is introduced in Section II. In Section III, the performance analysis of the average waiting time of packets for different scenarios in terms of multiple PU and single PU. Then, simulation and numerical results are reported in Section IV. Finally, we draw conclusions in Section V. II. SYSTEM MODEL Clearly, the sizes of packets used to carry the payload differ depending on the particular service. For example, voice services are rather sensitive to delay and the related packets are small in size to facilitate transmission in short time. On the other hand, short message services (SMS) are more tolerant to delay and hence longer SMS packets can be utilized. Given the above rationale, we assume that access to the transmission channel is divided into equal time slots and each user has two types of packets, namely, voice and data, representing delay sensitive and delay insensitive services. The related packets are stored in queues of infinite length and processed in first-in-first-out (FIFO) order. Besides the priority among users and differentiation between PU and SU, we also consider the priority for the two service types where voice shall be prioritized over data. Without loss of generality, we assume that the size of voice packets is shorter than the size of data packets and fulfill the condition L 0 <L 1 and L 2 <L 3 , where L 0 and L 1 are the sizes of voice and data packets for PUs, respectively, and L 2 and L 3 denote the sizes of voice and data packets for the SU, respectively. In the sequel, we consider two scenarios that are used to investigate the impact of multiple PUs and alternatively a single PU on the average waiting time for SUs. 2010 5th International Symposium on Wireless Pervasive Computing (ISWPC) 978-1-4244-6857-7/10/$26.00 ©2010 IEEE 122