Smart offloading for IoT application: Building a fog-cloud based context aware offloading framework and exploring potential for integration with blockchain

Abstract

In the world of interconnected devices also referred to as the Internet of Things (IoT) in the modern era, it's important to ensure that computing resources are allocated efficiently to nearby devices such as edge, fog, or cloud systems to meet resource needs. However, problems such as delays in data transmission, high energy consumption, and slow response times can negatively impact the performance of time-sensitive applications in cloud-based environments.

This paper presents the Context-Aware Offloading Framework (CAOF) for resource-constrained IoT applications. CAOF leverages contextual information to identify scenarios where offloading tasks to the cloud or to the local instances are beneficial. The framework aims to make optimal offloading decisions to improve system performance and minimize energy consumption. The effectiveness of CAOF is evaluated through simulations, comparing its performance against established offloading frameworks. CAOF is implemented as a middleware solution within an Amazon Web Services (AWS) ecosystem. This middleware integrates a Greengrass intelligent gateway that dynamically determines how to handle incoming data based on contextual information. The intelligent gateway can either process the data on local Elastic Cloud Compute (EC2) instances, effectively creating a fog layer, or send it directly to the cloud for further processing.

Experimental results demonstrate that CAOF achieves an energy consumption of 0.0011 joules approximately, with an memory utilization of 3.46 MB calculated as and average over all the EC2 machines. The framework execution time, averaging 4.07 s on edge, 5.41 s on cloud, and only 0.56 s when leveraging EC2 instances specifically, including an 80.4% accuracy in multi-class classification tasks. The CAOF systematically selects the most suitable alternatives for each offloading scenario to optimize efficiency in terms of time, memory, CPU, and energy consumption. The proposed smart gateway framework utilizes a hybrid approach to make optimal offloading decisions by considering contextual data. The research concludes with the design and development of an edge or fog-based framework that uses smart computing to make decisions using machine learning reasoning. The proposed framework architecture incorporates feature selection, classification, and hybrid logistic regression-based learning for the most effective offloading solution.