A Novel Service Deployment Policy in Fog Computing Considering The Degree of Availability and Fog Landscape Utilization Using Multiobjective Evolutionary Algorithms

TL;DR

This paper introduces a multiobjective evolutionary algorithm approach for optimizing service deployment in fog computing, balancing availability requirements and resource utilization, with MOPSO showing superior performance.

Contribution

It proposes a linear model for service placement considering availability and resource use, and compares three evolutionary algorithms for optimal deployment in fog environments.

Findings

MOPSO outperforms NSGA-II and MOEA/D in objective values and deadline satisfaction.

All three algorithms demonstrate efficiency in the fog computing context.

The approach effectively balances availability and resource utilization.

Abstract

Fog computing is a promising paradigm for real-time and mission-critical Internet of Things (IoT) applications. Regarding the high distribution, heterogeneity, and limitation of fog resources, applications should be placed in a distributed manner to fully utilize these resources. In this paper, we propose a linear formulation for assuring the different availability requirements of application services while maximizing the utilization of fog resources. We also compare three multiobjective evolutionary algorithms, namely MOPSO, NSGA-II, and MOEA/D for a trade-off between the mentioned optimization goals. The evaluation results in the iFogSim simulator demonstrate the efficiency of all three algorithms and a generally better behavior of MOPSO algorithm in terms of obtained objective values, application deadline satisfaction, and execution time.