Cost Efficient Resource Management in Fog Computing Supported Medical Cyber-Physical System

TLDR

Smart devices increasingly permeate daily life, and medical cyber‑physical systems (MCPSs) enable intelligent interaction between computational elements and medical devices, but cloud‑based processing suffers from unstable, high‑latency links, prompting the adoption of fog computing at the network edge to support MCPSs. This work proposes a fog‑computing‑supported MCPS (FC‑MCPS) that integrates fog resources with MCPS to achieve cost‑efficient operation. The authors jointly optimize base‑station association, task distribution, and virtual‑machine placement by formulating a mixed‑integer nonlinear program, linearizing it to a mixed‑integer linear program, and then solving it with a two‑phase LP‑based heuristic. Experiments demonstrate that the heuristic achieves near‑optimal cost efficiency and significantly outperforms a greedy baseline.

Abstract

With the recent development in information and communication technology, more and more smart devices penetrate into people's daily life to promote the life quality. As a growing healthcare trend, medical cyber-physical systems (MCPSs) enable seamless and intelligent interaction between the computational elements and the medical devices. To support MCPSs, cloud resources are usually explored to process the sensing data from medical devices. However, the high quality-of-service of MCPS challenges the unstable and long-delay links between cloud data center and medical devices. To combat this issue, mobile edge cloud computing, or fog computing, which pushes the computation resources onto the network edge (e.g., cellular base stations), emerges as a promising solution. We are thus motivated to integrate fog computation and MCPS to build fog computing supported MCPS (FC-MCPS). In particular, we jointly investigate base station association, task distribution, and virtual machine placement toward cost-efficient FC-MCPS. We first formulate the problem into a mixed-integer non-linear linear program and then linearize it into a mixed integer linear programming (LP). To address the computation complexity, we further propose an LP-based two-phase heuristic algorithm. Extensive experiment results validate the high-cost efficiency of our algorithm by the fact that it produces near optimal solution and significantly outperforms a greedy algorithm.