A forecasting-based approach for optimal deployment of edge servers in 5G networks

Vaibhav Tiwari,Chandrasen Pandey,Diptendu Sinha Roy

DOI: https://doi.org/10.1007/s10586-023-04250-0

2024-02-08

Cluster Computing

Abstract:The growing need for applications with low latency and high bandwidth in 5G networks has shifted the emphasis towards integrating edge computing into mobile network architectures. The optimal deployment of edge servers (ESs) is crucial for balancing performance and cost. Current research primarily focuses on the optimal deployment of ESs. However, there is limited attention on forecasting the workload of ESs, particularly in scenarios where they are placed in their optimal positions, which is crucial for accommodating anticipated variations in user densities and workload fluctuations expected in the near future. This research forecasting of workload in 5G networks by leveraging the Temporal Hierarchical Attention Mechanism Network (THAMNET) model for accurate Internet traffic forecasting. Addressing this gap, the Max–Min Fairness Allocation Scheme (MMF-AS) algorithm is implemented to ensure a balanced workload distribution. This approach facilitates fair allocation of Base Stations (BSs) to ESs, taking into account factors such as workload, distance, and connectivity. A novel approach is introduced by integrating the THAMNET forecasting model with the Particle Swarm Optimization (PSO) and MMF-AS algorithms. The proposed method considers predicted future traffic demands, and connectivity between ESs and BSs, and aims to achieve balanced workload distribution, higher utilisation rates, minimised latency, and reduced energy consumption. Experimental results demonstrate equitable workload distribution 25.95%, utilisation rate 17.57%, and notable reductions in latency 47.61% and energy consumption 32.28% compared to the non-forecasting comparison.

computer science, information systems, theory & methods