EcoTaskSched: a hybrid machine learning approach for energy-efficient task scheduling in IoT-based fog-cloud...

Area of Science:

• Computer Science
• Artificial Intelligence
• Cloud Computing

Background:

• Cloud computing adoption presents energy and resource efficiency challenges.
• Fog-cloud computing integrates resources to address these issues, but optimal task allocation and energy management remain complex.
• Efficient utilization of heterogeneous fog-cloud resources for energy-efficient task scheduling is a critical research problem.

Purpose of the Study:

• To propose a novel Machine Learning (ML)-based model, EcoTaskSched, for energy-efficient task scheduling in fog-cloud networks.
• To leverage deep learning, specifically Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM), for optimized task placement and energy reduction.
• To ensure Quality of Service (QoS) provisioning while enhancing energy efficiency in fog-cloud environments.

Main Methods:

• Developed a hybrid ML model (EcoTaskSched) integrating CNNs for feature extraction and BiLSTM for sequential information processing.
• Implemented a real fog-cloud environment using the COSCO framework and Azure B2s nodes for simulation.
• Utilized the DeFog benchmark for task workloads, with data preprocessing including normalization, feature engineering, and augmentation.

Main Results:

• EcoTaskSched demonstrated significant reductions in energy consumption and improvements in job completion rates compared to baseline models.
• Achieved an 85% job completion rate, outperforming GGCN and BiGGCN, with lower average response times and SLA violation rates.
• Showcased increased throughput and reduced execution costs, confirming its effectiveness in optimizing fog-cloud task scheduling.

Conclusions:

• The EcoTaskSched model effectively enhances task handling efficiency and reduces energy consumption in fog-cloud computing environments while maintaining QoS.
• The model's performance indicates its successful application in optimizing scheduling algorithms for diverse fog-cloud settings.
• Future work will involve long-term testing in real-world IoT environments and exploring integration with other ML models for further optimization.