Enhancing modular application placement in a hierarchical fog computing: A latency and communication cost-sensitive approach

Over the years, cloud computing has been a key enabler for handling complex applications and services for Internet of Things (IoT) devices situated at the edge of the network. Services and applications that are driven by the IoT environment commonly have stringent latency-sensitive requirements and may experience long network delays due to the long physical distance of cloud-based computational resources from IoT devices. Fog computing gained adoption as a solution in this case because it shortens this distance by spreading the computing power around the edge of the network in tiers. This contributes to network latency reduction and response times improvements of applications that have sensitive temporal requirements, besides improving the overall data traffic management in the network. Nevertheless, when certain requirements of an application are prioritized over others during the resource allocation process, a fog tier closer to IoT devices may experience resource depletion, forcing other latency-sensitive applications to use resources from a distant fog level and causing them to become non-responsive. To address this issue, this work proposes an approach for allocating modular applications in a hierarchical tier-based fog computing architecture. The proposed approach, named Least Impact - X (LI-X), aims to minimize the response time of latency-sensitive applications and reduce data traffic on the network by mitigating the idle time of resources at the lower levels of the hierarchical fog. This is achieved by distributing the application modules among the fog tiers in order to minimize the response time of delay-sensitive applications, while also reducing the overall network traffic. The performance of LI-X was compared to previous studies in a simulated iFogSim environment. Results have demonstrated that LI-X outperforms these studies in most of the proposed scenarios, effectively reducing response time and minimizing communication data costs on the network.