On the Fog-Cloud Cooperation: How Fog Computing can address latency concerns of IoT applications

Fog computing emerged as a new computing paradigm which moves the computing power to the proximity of users, from core to the edge of the network. It is known as the extension of Cloud computing and it offers inordinate opportunities for real-time and latency-sensitive IoT applications. An IoT application consists of a set of dependent Processing Elements (PEs) defined as operations performed on data streams and can be modeled as a Directed Acyclic Graph (DAG). Each PE performs a variety of low-level computation on the incoming data such as aggregation or filtering. A key challenge is to decide how to distribute such PEs over the resources, in order to minimize the overall response time of the entire PE graph. This problem is known as distributed PE scheduling and placement problem. In this work, we try to address the question of how fog computing paradigm can help reducing the IoT application response time by efficiently distributing PE graphs over the Fog-Cloud continuum. We mathematically formulate the fundamental characteristics of IoT application and Fog infrastructure, then model the system as an optimization problem using Gravitational Search Algorithm (GSA) meta-heuristic technique. Our proposed GSA model is evaluated by comparing it with a well-known evolutionary algorithm in the literature via simulation. Also, a comparative analysis with the legacy cloud infrastructure is done in order to show the significant impact of fog presence on the performance of PE processing. Evaluation of our model demonstrates the efficiency of our approach comparing to the current literature.