On-demand Vehicular Fog Computing for Beyond 5G Networks

• Published in: IEEE transactions on vehicular technology 2023-12, Vol.72 (12), p.1-17
• Main Authors: Mao, Wencan, Akgul, Ozgur Umut, Cho, Byungjin, Xiao, Yu, Yla-Jaaski, Antti
• Format: Article
• Language: English
• Subjects: and techno-economic analysis; Capacity planning; Cloud computing; Computational modeling; Computing costs; Costs; Economic analysis; Economic models; Edge computing; integer linear programming (ILP); Integer programming; Linear programming; Network latency; Nodes; Optimization; Quality of service; quality of service (QoS); Regional planning; Resource management; Revenue; Routing; Traffic flow; Vehicle dynamics; vehicular fog computing (VFC)
• ISSN: 0018-9545; 1939-9359
• DOI: 10.1109/TVT.2023.3289862

Emerging compute-intensive and latency-sensitive vehicular applications are expected to be deployed at the edge instead of the cloud to shorten the network latency. Mobile fog nodes carried by moving vehicles, namely vehicular fog nodes (VFNs), have been proposed to complement the stationary fog nodes co-located with base stations to handle the spatiotemporal variations of demand in a cost-efficient way. Existing works on capacity planning for such vehicular fog computing (VFC) scenarios are built on the assumption of certain spatiotemporal patterns of vehicular traffic. They consider long-term capacity planning (e.g., updated every season) but leave the adaptation to temporary changes or unexpected variations out of scope. These solutions typically result in high computational costs and thus are not suitable for short-term capacity planning, which requires low-latency responses. To reduce time complexity, we propose an integer linear programming (ILP)-based framework called on-demand capacity planning (ODCP) to implement two-phase planning through optimizing the routing strategies of VFNs, with the aim of maximizing the profit and quality of service (QoS). More specifically, ODCP first predicts the traffic flow and resource demand using seasonal autoregressive integrated moving average (SARIMA) and estimates the revenue using an economic model defined by service level agreement (SLA). With the estimated workload and revenue, the first phase (i.e., global planning) decides the ratio of tasks that can be served at the city scale and assigns VFNs to each region. The second phase (i.e., regional planning) assigns the VFNs to users within the same region and schedules the routes of VFNs based on the mobility of users. Experimental results show that the proposed solution achieves a higher performance in terms of profit and QoS than the existing single-phase capacity planning solutions. We also find that a large number of VFNs, a small region size, high penalty costs, and low travel and rental costs lead to high service rates, whereas a large region size and low travel, rental, and penalty costs lead to high profits.