Low Power Scheduling Approach for Heterogeneous System Based on Heuristic and Greedy Method.

Area of Science:

• Computer Science
• Artificial Intelligence
• Cloud Computing

Background:

• Big data, AI, and cloud computing in heterogeneous systems raise energy consumption concerns.
• High energy usage impacts operational costs and system reliability.
• Addressing energy efficiency is critical for system architects and researchers.

Purpose of the Study:

• To propose a novel energy-saving task scheduling method for heterogeneous systems.
• To reduce energy consumption while maintaining or improving system performance.
• To offer an alternative to traditional 0-1 programming for task allocation.

Main Methods:

• Developed a heuristic and greedy energy saving (HGES) approach.
• Tasks are initially assigned to all GPUs, then categorized into high-value and low-value based on time value and variance.
• Employed a greedy strategy to schedule high-value tasks first, followed by low-value tasks.

Main Results:

• The HGES method demonstrated superior energy saving compared to existing methods.
• HGES achieved faster results than 0-1 programming.
• Experimental validation on diverse platforms confirmed the method's effectiveness and rationality.

Conclusions:

• The HGES approach provides an effective solution for energy saving in heterogeneous computing environments.
• This method offers a faster and more energy-efficient alternative for task scheduling.
• The findings support the adoption of HGES for optimizing resource utilization and reducing operational costs.