Hybrid Fog Resource Provisioning and Resource Allocation using MKFCM and FPA

Abstract

The expansion of Internet of Things (IoT) deployments has motivated the Fog Computing (FC) paradigm, which brings mobile-edge and cloud resources together under a single model. Allocating and provisioning application tasks across such environments poses challenges due to constrained resource capabilities, heterogeneity, mobility, network hierarchy, and unpredictability. This paper addresses the need for additional computing devices that prioritize delay-sensitive applications by proposing a Hybrid Fog Resource Provisioning (HFRP) Algorithm for optimal resource allocation. The HFRP Algorithm combines Modified K-means Fuzzy CMeans (MKFCM) clustering and the Flower Pollination Algorithm (FPA) to efficiently select and provision resources. The algorithm operates in two phases: (i) Resource Discovery and Clustering, which identifies available resources and clusters them using MKFCM and (ii) Resource Allocation, which utilizes FPA to allocate tasks to the clustered resources. Simulation results exhibit that the HFRP algorithm significantly supersedes several other algorithms, including WORA, PSO, GA, and SJF. Key performance improvements include: HFRP reduces average cost by 25.83% compared to WORA, 33.7% compared to PSO, 45.9% compared to GA and 64.7% compared to SJF. HFRP consumes 16.69% less energy than WORA, 32.3% lower consumption than PSO, 38.7% lower consumption than GA and 59.77% lower than SJF. HFRP achieves 7.2% better makespan than WORA, 9.1% improvement over PSO, 15.8% improvement over GA, and 30.4% better than SJF. Successful Task Completion Ratio: HFRP surpasses WORA by 1.97%, 11.28% over PSO, 10.29% over GA and 27.12% over SJF. The proposed HFRP algorithm efficaciously addresses the challenges of resource allocation and provisioning in FC environments. By leveraging MKFCM for resource clustering and FPA for task allocation, HFRP achieves significant improvements in cost reduction, energy consumption, average loop delay, makespan, and task completion ratio compared to existing algorithms. These results bring out the potential of HFRP for augmenting the performance and efficiency of IoT applications in FC environments.