Analysis and design of scheduling schemes for wireless networks with unsaturated traffic

• Published in: China communications Vol. 18; no. 2; pp. 65 - 85
• Main Authors: Li, Jianfei, Wen, Juan, Sheng, Min
• Format: Journal Article
• Language: English
• Published: China Institute of Communications 01.02.2021; State Key Laboratory of Integrated Services Networks, Xidian University, Xi'an 710071, China
• Subjects: Delays; markov chain; performance analysis; Quality of service; Relays; Schedules; Scheduling; Scheduling algorithms; Throughput; unsaturated traffic; unsaturated traffic; scheduling; perfor-mance analysis; markov chain
• ISSN: 1673-5447
• DOI: 10.23919/JCC.2021.02.006

Scheduling schemes assign limited resources to appropriate users, which are critical for wireless network performance. Most current schemes have been designed based on saturated traffic, i.e., assuming users in networks always have data to transmit. However, the user buffer may sometimes be empty in actual network. Therefore, these algorithms will allocate resources to users having no data to transmit, which results in resource waste. In view of this, we propose new scheduling schemes for one-hop and two-hop link scenario with unsaturated traffic. Furthermore, this paper analyzes their key network performance indicators, including the average queue length, average throughput, average delay and outage probability. The two scheduling algorithms avoid scheduling the links whose buffers are empty and thus improve the network resource utilization. For the one-hop link scenario, network provides differentiated services via adjusting the scheduling probabilities of the destination nodes (DNs) with different priorities. Among the DNs with same priority, the node with higher data arrival rate has larger scheduling probability. For the two-hop link scenario, we prioritize the scheduling of relay-to-destination (R-D) link and dynamically adjust the transmission probability of source-to-relay (S-R) link, according to the length of remaining buffer. The experiment results show the effectiveness and advantage of the proposed algorithms.