ML-Guided Application Energy Tuning for Computing Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large organizations face challenges in efficiently managing the electrical energy consumption of complex computing systems, leading to increased environmental impact and operational inefficiencies.
Innovation Solution
A system and method that utilizes machine learning to analyze energy use data, suggesting changes to applications to reduce energy consumption, thereby optimizing computational resources and potentially reducing the number of devices and data centers needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If computing devices perform more operations to maintain or improve speed, then application performance is improved, but electrical energy consumption increases
Solution Approach 1:
The system changes operational parameters by analyzing performance data and energy use data to identify optimal operating points. Machine learning models determine suggested changes to application operations that maintain performance while reducing energy consumption, effectively changing the parameters of computation to achieve better efficiency.
Solution Approach 2:
The system implements continuous feedback by monitoring both performance data and energy use data from computing devices. This feedback loop enables the machine learning models to learn from actual operational patterns and suggest adjustments that balance performance requirements with energy consumption goals.
2Loss of energy
If the number of computational devices is reduced, then energy consumption and environmental impact are reduced, but the capability to perform operations may be compromised
Solution Approach 1:
The system enables computing devices to self-optimize by automatically analyzing their own performance and energy data. The machine learning models generate suggested changes that devices can implement autonomously to reduce energy consumption while maintaining required computational capabilities, eliminating the need for additional devices to compensate for efficiency losses.
Solution Approach 2:
The system creates a universal machine learning model that can be applied across different application types and computing devices. By training on diverse energy use data from multiple applications, the model learns general optimization strategies that can be transferred to reduce energy consumption across various computational workloads without requiring device-specific customization.
Data Source
AI summary
A system for improving energy use associated with an application includes a processor configured to receive performance data from a device associated with the application for a first period of time and updates energy-use data stored in memory with the received performance data. The processor then determines an energy use value based at least in part upon the energy use data stored in the memory. When the energy use value is greater than a predetermined threshold, the processor analyzes the energy use data using machine learning, which produces suggested changes to the application that are determined by machine learning to reduce the energy use value. The processor then initiates the suggested changes by sending the suggested changes to the device associated with the application.