Task offloading using fog analytics
Rakesh Matam, Somanath Tripathy · 2024
Fog computing has recently evolved as a promising technology to meet the specific latency and QoS requirements of Internet of Things (IoT) applications. End-devices use fog computing facilities to minimize latency, communication costs, and energy in the cloud-based frameworks. This, however, has many associated challenges, mainly due to the distributed nature of the fog network. As different applications have different QoS requirements, a single framework for task offloading would not be optimal. Thus, the issues in task offloading include the decisions of when, where, and what tasks to offload, as well as how the offloading task be scheduled to minimize latency. Besides that, limited energy, device heterogeneity, and task workload would influence the offloading decision. In view of different application requirements, the objective of task offloading becomes a multi-objective optimization problem. In this work, we first present the impact of various parameters on task-offloading decisions. Subsequently, we analyze the impact of different fog analytics schemes and learning algorithms on task offloading. This study presents the optimal performance criterion for task-offloading decisions using fog analytics. The data collected from multiple fog-nodes at discrete time intervals in a distributed manner is used to select a fog node for task offloading.