ARMA-Prediction-Based Online Adaptive Dynamic Resource Allocation in Wireless Virtualized Network

• Main Authors: Lun Tang, Xiaoyu He, Xixi Yang, Yannan Wei, Xiao Wang, Qianbin Chen
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Wireless virtualized networks; resource allocation; multiple time-scales; ARMA; large-deviation principle
• Online Access: https://ieeexplore.ieee.org/document/8832122/

Wireless network virtualization (WNV) provides a novel paradigm shift in the fifth-generation (5G) system, which enables to utilize network resources more efficiently. In this paper, by jointly considering cache space and time-frequency resource allocation in wireless virtualized networks, we first formulate an optimization programming to investigate the minimization problem of network overheads while satisfying the quality of service (QoS) requirements of each virtual network on overflow probability. Then, with diverse demands of virtual networks for different kinds of resources taken into consideration, an online adaptive virtual resource allocation algorithm with multiple time-scales based on auto regressive moving average (ARMA) prediction method is proposed to solve the formulation, which could eliminate the irrationalities existed in traditional approaches caused by the uncertainty of traffic and information feedback delay. More specifically, in the proposed resource scheduling mechanism with multiple time-scales, on the one hand, a reservation strategy of cache space is developed according to the ARMA prediction information under long time-scales. On the other hand, virtual networks are sorted by the overflow probabilities derived by the large-deviation principle and dynamic time-frequency resource scheduling under short time-scales. Simulation results reveal that our proposal can provide tangible gains in reducing the bit loss rate and improving the utilization of physical resources.