Optimizing edge server placement and load distribution in mobile edge computing using ACO and heuristic algorithms
Sevda Zarei,Sadoon Azizi,Awder Ahmed
DOI: https://doi.org/10.1007/s11227-024-06780-9
IF: 3.3
2024-12-09
The Journal of Supercomputing
Abstract:With the rapid growth and development of Internet of Things (IoT) and smart mobile devices, the volume of data generated at the network edge has increased significantly. Mobile edge computing (MEC) has emerged as a pivotal technology to address the computational limitations of these devices by bringing cloud capabilities closer to end users. However, MEC still faces two critical challenges: edge server placement (ESP) and load distribution. The first problem involves determining the optimal locations for a set of heterogeneous computational servers within a given network of base stations (BSs) and their associated workload. The objective is to strategically position these servers to ensure efficient computational resource allocation and to enhance the quality of service (QoS) for end users. The second problem is the distribution of computational tasks/requests, which itself comprises two sub-problems: (a) identifying the most suitable server for the requests arriving at each BS, and (b) scheduling these requests on the selected servers. This paper addresses these challenges through a comprehensive approach. We model these problems using a mixed-integer nonlinear programming (MINLP) framework. For the ESP problem, the goal is to achieve load balancing among servers. For the load distribution problem, the objectives are to minimize response time and ensure that user request deadlines are met. To address these problems, we propose an ant colony optimization (ACO) algorithm to efficiently solve the ESP problem, ensuring balanced load among servers. For the load distribution problem, we introduce two heuristic algorithms: one for selecting the most suitable server to distribute incoming workloads from devices, and another for scheduling requests on each server based on their urgency and importance. We validate our proposed approach through extensive experiments using real-world data from Shanghai Telecom base stations. The experimental results demonstrate that our approach significantly outperforms state-of-the-art methods in terms of load standard deviation, average request response time, and the percentage of deadlines successfully met. Specifically, our method improves load balancing by 57.6%, reduces the average response time by 50.7%, and increases deadline compliance by 35.2%. These findings underscore the efficacy of our proposed methods in enhancing MEC system performance, thereby delivering superior QoS and user experience in IoT and mobile computing applications.
computer science, theory & methods,engineering, electrical & electronic, hardware & architecture