Software-Based Computer Process Energy Prediction Without External Hardware
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately measure and predict the energy consumption of computer processes across various computing platforms and hardware without requiring external hardware, while being agnostic to process and platform specifics, and efficiently separate process energy consumption from that of other processes.
Innovation Solution
A framework that measures resource utilization and energy consumption using software, creates a labeled dataset, builds a machine learning prediction model, and applies it to predict energy consumption of specific processes, treating them as 'black-boxes', applicable to both Windows and Linux systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external hardware is used to measure energy consumption, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts the energy measurement function from external hardware and implements it through software-based resource utilization monitoring. The system measures energy consumption by tracking CPU, memory, storage, and network resource usage of individual processes, eliminating the need for external power measurement devices while maintaining measurement capability through operational data correlation
Solution Approach 2:
The patent replaces the mechanical/physical measurement system (external hardware power meters) with a software-based information processing system. Instead of directly measuring electrical power consumption through hardware, the system uses software to monitor resource utilization metrics and correlates these with energy consumption through machine learning models, substituting physical measurement with computational analysis
2Measurement precision
If process-specific energy measurement is implemented, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the overall energy consumption measurement into process-specific components by monitoring individual process resource utilization separately. The system tracks CPU time, memory usage, storage I/O, and network activity for each process independently, then uses machine learning to attribute total energy consumption to individual processes based on their resource usage patterns, enabling precise process-level measurement without requiring separate physical sensors for each process
Solution Approach 2:
The patent changes the measurement parameters from direct physical power consumption to operational resource utilization metrics. By monitoring software-accessible parameters such as CPU cycles consumed, memory pages accessed, disk I/O operations, and network packets transmitted, the system transforms energy measurement into a software-based parameter observation problem that can be solved through correlation analysis and machine learning
3Measurement precision
If comprehensive resource utilization measurement is performed, then prediction accuracy is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary action by pre-training machine learning models using comprehensive resource utilization data collected during system operation. The system gathers training data from multiple processes across different workloads, trains the energy prediction models in advance, and stores the learned relationships between resource usage and energy consumption. This allows the system to make rapid predictions without performing time-consuming real-time analysis of all resource metrics
Solution Approach 2:
The patent uses copying by creating a virtual model of the energy consumption system through machine learning. Instead of performing direct real-time measurement and analysis of all resource parameters, the system creates a trained predictive model that copies the energy consumption behavior patterns. This model can then rapidly predict energy usage by simply observing current resource utilization, avoiding the time cost of comprehensive real-time analysis
Data Source
Figure 1
AI summary
A method for measuring and predicting the energy consumption of computer processes running on a computerized device, comprising the steps of measuring the resource utilization and the corresponding energy consumption of the utilization, for a wide variety of resource usage combinations; creating a labeled dataset, which consists of the measurements obtained in the preceding step; building a machine learning prediction model that predicts the energy consumption of any resource utilization combination, using the labeled dataset; monitoring the resource utilization of a specific process of interest; predicting the energy consumption of this specific process in a predetermined context, by applying the prediction model.