Adaptive Packet-size Control for Improved Throughput in Dynamic Access Networks

Abstract

Cognitive radio (CR) is a new intelligent wireless technology that aims at improving spectrum utilization by allowing opportunistic access to the underutilized licensed spectrum. Wireless CR operating environment is typically characterized by its unreliable and unpredictable channel conditions and time availability due to fading and the randomness of primary radio (PR) activities. In such environment, packet fragmentation is needed to enhance the probability of success/packet delivery and reduce the needed number of packet re-transmission attempts. Specifically, the quality and availability of the PR channels along with the data packet size should be considered when designing communication protocols for CR networks (CRNs) such that the packet success probability is improved. Based on the quality and availability of the PR channels, an optimal-packet size metric is derived, which is defined as the packet size that can be transmitted over a selected channel while guarantying a predefined probability of success. In this paper, we propose three fragmentation-based channel assignment algorithms: fixed-fragment size algorithm, first-fit algorithm and near-exact fit algorithm. The first algorithm divides the packet equally over the selected channels while the other algorithms use variable fragment size. The main objectives of the algorithms are to enhance network throughput, decrease the number of dropped packets and reduce the average number of retransmission attempts. This preserves more channels for potential future CR transmissions, resulting in higher network throughput with less energy consumption. Simulation results show that the proposed algorithms significantly outperform existing CRN channel assignment algorithms with no fragmentation.