Hybrid-grained migration method for load redistribution in heavily-loaded fog nodes

Abstract

Fog paradigms minimize response times for delay-sensitive applications by bringing computation and storage closer to the edge of the network, thus reducing the need for data transmission to centralized cloud servers. However, one key challenge of fog computing is the constrained network resources in comparison with the cloud core. This yields a bottleneck when the delay-sensitive applications demand intensive computation and processing resources. Here migrating tasks to the cloud can be prohibited due to the delay bounds impelled by offered services. Therefore, running delay-sensitive and computation-intensive applications at the fog nodes compels efficient resource management solutions to avoid saturation and congestion. Along this, this paper presents a novel dynamic redistribution mechanism that alleviates heavily-loaded (HL) fog nodes from excess load and instead migrates it to nearby lightly-loaded (LL) nodes. The method implements a hybrid model that combines the saliencies of coarse-grained and fine-grained techniques, i.e., adaptive selection based on the load status at the proximity nodes and traffic volume. Results demonstrate that the proposed method results in minimal migration iterations at low failure rates, along with reduced migration delay and cost, as compared to standalone fine-grained and coarse-grained methods.