An Energy-Friendly Scheduler for Edge Computing Systems

Area of Science:

• Edge Computing
• Resource Management
• Sustainable Computing

Background:

• Modern applications like the Internet of Things (IoT) demand high availability, low latency, and low energy consumption.
• Edge computing, utilizing single board computer (SBC) clusters near users, addresses these demands but faces challenges with battery-powered deployments.
• Container orchestration platforms like Kubernetes manage services but lack energy-aware placement strategies.

Purpose of the Study:

• To develop an optimized scheduler for Kubernetes within energy-constrained SBC clusters.
• To improve resource management for battery-powered edge devices.
• To enhance the efficiency and reliability of edge computing services.

Main Methods:

• Proposed a novel energy-aware scheduler for Kubernetes.
• Integrated the scheduler into single board computer (SBC) clusters.
• Evaluated scheduler performance against existing solutions in energy-constrained environments.

Main Results:

• Achieved a 23% reduction in event rejections.
• Reduced deadline violations by 83%.
• Demonstrated an approximate 59% decrease in overall cluster energy consumption.

Conclusions:

• The proposed energy-aware scheduler significantly enhances the performance and energy efficiency of Kubernetes on SBC clusters.
• This approach is crucial for sustainable and reliable operation of battery-powered edge computing applications.
• Optimized scheduling addresses key limitations of current orchestration platforms in energy-constrained scenarios.