Maximizing Revenue for Reliability-Aware Edge Application Deployment

Multiaccess edge computing (MEC) enables low-latency service delivery by deploying application instances on edge servers. However, edge servers are prone to failures, making it challenging to meet diverse user reliability requirements. A common approach is to deploy redundant instances across multiple edge servers, which improves reliability but increases costs and limits the number of users that can be served within budget. Therefore, efficient deployment strategies are needed to balance cost-effectiveness and reliability guarantees, thereby maximizing the app vendor’s revenue. In this article, we investigate the problem ofRevenue maximization forReliability-awareEdgeApplicationDeployment ($\text{R}^{2}\text{EAD}$). Our objective is to maximize the app vendor’s revenue by deploying its applications on heterogeneous edge servers, subject to budget and resource constraints and users’ diverse reliability requirements. We prove that the$\text{R}^{2}\text{EAD}$problem is$\mathcal {\text{NP}}$-hard and propose an efficient approximation algorithm named$\text{R}^{2}\text{EAD}$-A. By reducing the problem to a nonmonotone submodular maximization problem with curvature$\alpha$under multiple knapsack constraints, we prove that$\text{R}^{2}\text{EAD}$-A achieves a constant approximation ratio of$\frac{1}{\alpha }(1 - e^{-\alpha })$. Extensive evaluations demonstrate that$\text{R}^{2}\text{EAD}$-A outperforms the representative approaches across all tested cases.