Mobility-Aware Task Offloading and Migration Schemes in Fog Computing Networks

Abstract

Fog computing is an extension of cloud computing, which emphasizes distributed computing and provides computing service closer to user equipments (UEs). However, due to the limited service coverage of fog computing nodes (FCNs), the moving users may be out of the coverage, which would cause the radio handover and execution results migration when the tasks are off-loaded to FCNs. Furthermore, extra cost, including energy consumption and latency, is generated and affects the revenue of UEs. Previous works rarely consider the mobility of UEs in fog computing networks. In this paper, a generic three-layer fog computing networks architecture is considered, and the mobility of UEs is characterized by the sojourn time in each coverage of FCNs, which follows the exponential distribution. To maximize the revenue of UEs, the off-loading decisions and computation resource allocation are jointly optimized to reduce the probability of migration. The problem is modeled as a mixed integer nonlinear programming (MINLP) problem, which is NP-hard. The problem is divided into two parts: tasks off-loading and resource allocation. A Gini coefficient-based FCNs selection algorithm (GCFSA) is proposed to get a sub-optimal off-loading strategy, and a distributed resource optimization algorithm based on genetic algorithm (ROAGA) is implemented to solve the computation resource allocation problem. The proposed algorithms can handle the scenario of UEs' mobility in fog computing networks by significantly reducing the probability of migration. Simulations demonstrate that the proposed algorithms can achieve quasi-optimal revenue performance compared with other baseline algorithms.

Keywords

• Fog computing
• mobility
• Gini coefficient
• resource optimization
• utility function