Cost-Efficient Design for 5G-Enabled MEC Servers under Uncertain User Demands

TL;DR

This paper presents a stochastic optimization framework for designing cost-efficient 5G-enabled MEC servers that adapt to uncertain user demands, minimizing capacity and latency.

Contribution

It introduces a novel accelerated Benders decomposition method to efficiently solve large-scale MILP models for MEC server planning under demand uncertainty.

Findings

ABD achieves optimal solutions with reduced computation time.

The model effectively balances capacity minimization and latency reduction.

Numerical results validate the approach's scalability and efficiency.

Abstract

Mobile edge computing (MEC) enhances the performance of 5G networks by enabling low-latency, high-speed services through deploying data units of the base station on edge servers located near mobile users. However, determining the optimal capacity of these servers while dynamically offloading tasks and allocating computing resources to meet uncertain user demands presents significant challenges. This paper focuses on the design and planning of edge servers with the dual objectives of minimizing capacity requirements and reducing service latency for 5G services. To handle the complexity of uncertain user demands, we formulate the problem as a two-stage stochastic model, which can be linearized into a mixed-integer linear programming (MILP) problem. We propose a novel approach called accelerated Benders decomposition (ABD) to solve the problem at a large network scale. Numerical experiments demonstrate that ABD achieves the optimal solution of MILP while significantly reducing computation time.