Dynamic Network Slice Scaling Assisted by Prediction in 5G Network

Abstract

Network slicing is a key technology in fifth-generation (5G) mobile networks. Slicing divides a physical network into multiple dedicated logical networks to meet the requirements of diverse use cases. Efficient slice deployment algorithms are critical in reducing network operators' costs and energy consumption and in providing users better service. Many researchers have focused on static deployment when investigating network slices, effectively ignoring network operators' requirements for the dynamic deployment and expansion of such slices. In this paper, we first construct a joint optimization problem of cost and energy consumption. Then, we propose a prediction-assisted adaptive network slice expansion algorithm to deploy network slices dynamically. The proposed algorithm consists of three parts. First, we devise a Holt-Winters (HW) prediction algorithm to determine traffic demand for network slices. This method is intended to avoid frequent changes in network topology. Second, we propose a virtual network function (VNF) adaptive scaling strategy to reasonably determine the number of VNFs and resources required for network slices to avoid resource wastage. Finally, we develop a proactive online algorithm to deploy network slices. This method deploys network slices reasonably via the VNF deployment algorithm and link-routing algorithm to ensure slices' service requirements. Resource capacity and delay requirements are also considered in our evaluation to ensure that network costs and energy consumption are minimized. We then perform a series of simulation experiments to compare the proposed method's performance to state-of-the-art dynamic network slicing technologies. Ultimately, our solution is deemed a suitable candidate for dynamic deployment of 5G network slices; the solution demonstrates advantages of high resource utilization, low deployment costs, and low energy consumption.