Centralized Framework for Workload Distribution in Fog Computing

Abstract

Fog computing is a platform for executing computation intensive tasks by aggregating computational resources of the edge-devices. The challenges of fog computing, includes heterogeneity, scalability, addressing and security. In this paper, we address the workload distribution problem in fog network. This is particularly important for reducing the computational latency of an offtoading task. An efficient workload distribution is required for effective resource utilization of the fog network. The two extreme approaches for solving the workload distribution problem is by utilizing a centralized or a distributed approach. A distributive framework can be scalable and fault tolerant. However, the communication overhead for maintaining a distributed framework is significantly more compared to a centralized approach. On the other hand, a centralized framework, can be relatively easier for implementing an economic model for multiple services offered by the system. In this paper, we investigate the feasibility of implementing a centralized framework to handle workload distribution in fog network. The two main entities of the framework are: Controller, that collects the available network resources of the participating devices for efficient workload distribution and Clients, providing computational resources for executing task. We evaluate our framework by integrating various heterogeneous nodes like DragonBoard, RaspberryPi and others.