A Comparative Study for Server Selection Schemes in Multiserver Mobile Edge Computing

The proliferation of resource-intensive applications (such as face recognition, image/video processing, interactive gaming, and augmented reality) on mobile devices poses challenges due to their high resource demands. While cloud offloading can alleviate some of these issues, it is unsuitable for latency-sensitive applications. Mobile Edge Computing (MEC) offers an alternative by offloading computational tasks to edge servers located at the periphery of mobile networks. However, selecting the optimal edge server remains complex due to limited resources, user mobility, and varying server workloads. In this study, we compare three edge server selection methodologies: Nearest Server (NS), Least Remaining CPU cycles (LRC), and Fewest Active Tasks (FAT). Using a central controller model, we evaluate these methods based on average task sojourn time, considering factors such as task size, CPU requirements, and communication/computation delays. Our findings provide insights into the effectiveness of each approach, contributing to a deeper understanding of edge server selection in dynamic MEC environments. Additionally, we propose a round-robin process scheduling model to enhance resource efficiency compared to traditional fixed CPU allocation methods. The results from the simulated experiments demonstrate that the Least Remaining CPU Cycles (LRC) scheme achieves the best overall performance in terms of task sojourn time, whereas the Nearest Server (NS) strategy performs the worst. Furthermore, the results underscore the trade-offs associated with different server selection strategies, providing valuable insights for optimizing MEC performance