A tri-objective model for cloudlet server placement problem in wireless metropolitan area networks

To reduce latency and save energy, cloudlet computing enables tasks to be offloaded from user equipment to Cloudlet Servers (CSs). Determining the optimal number of CSs and the appropriate locations for their placement are two major challenges in building an efficient computing platform. Placing a CS at the closest location to the user can improve the QoS. Additionally, providing additional CSs to cover each user ensures that the user's needs are met even if the designated server is unable to provide services. However, to minimize energy consumption and costs, service providers tend to use a minimum number of CSs. Since the coverage zones of different CSs may overlap, fewer additional servers need to be deployed in such areas. This paper examines the problem of CS placement in a Wireless Metropolitan Area Network (WMAN) and introduces a three-objective model that aims to optimize transmission distance, coverage with overlap control, and energy consumption. To obtain an appropriate Pareto front, the performance of the NSGA-II, binary MOPSO, and binary MOGWO algorithms is examined through four different scenarios under the Shanghai Telecom dataset. Comparing the results of the Hyper-Volume (HV) indicator reveals that the NSGA-II algorithm has higher values in all studied scenarios. A higher HV value means that the solution set is closer to an optimal Pareto set. In the best and worst case, the HV values for the NSGA-II were equal to 0.2275 and 0.1883, respectively.