ML Power Model for Request-Level Energy Attribution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current power consumption monitoring approaches in computing environments, such as data centers and cloud-based systems, primarily focus on infrastructure levels, lacking effective methods to measure or determine application-level energy consumption, especially in complex shared environments like database as a service (DBaaS).

Innovation Solution

A computer-implemented method is developed to train a machine learning power consumption model using a training dataset derived from historical request-related data, including request-related data and resource power consumption data. This model estimates power consumption by analyzing requests, identifying common patterns of resource use, and collecting real-time trace metrics to allocate resource use effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If infrastructure-level power consumption monitoring is implemented, then power consumption can be measured at the device level, but application-level power consumption cannot be determined

Engineering Contradiction:

Improvepower consumption measurement levelVSAvoidmonitoring system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments power consumption monitoring from infrastructure level to application level by introducing application performance monitoring agents that track individual application processes. These agents collect metrics specific to each application (CPU usage, memory access, I/O operations) and combine them with infrastructure power consumption data to calculate application-specific power consumption, thereby enabling precise measurement at the application level without requiring complete system redesign

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary component - the application performance monitoring agent - that bridges the gap between infrastructure monitoring and application-level measurement. This agent acts as a mediator by collecting application-specific metrics, correlating them with infrastructure power consumption data, and producing application-level power consumption estimates, thus resolving the limitation of traditional infrastructure-only monitoring

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If application-level power consumption monitoring is implemented, then energy usage can be tracked per application, but measurement accuracy decreases due to allocation complexity

Engineering Contradiction:

Improveapplication-level power consumption measurementVSAvoidpower consumption estimation accuracy

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent changes the parameters used for power consumption estimation by introducing application-specific performance metrics (CPU usage percentage, memory access patterns, disk I/O volume, network traffic) as weighting factors. Instead of using raw infrastructure power data alone, the system transforms these parameters into weighted contributions that reflect each application's actual resource utilization, thereby improving the accuracy and reliability of application-level power consumption estimates

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If detailed request-level analysis is performed, then power consumption can be attributed to individual requests, but processing time and complexity increase

Engineering Contradiction:

Improverequest-level power consumption attributionVSAvoidprocessing time for power consumption calculation

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by having application performance monitoring agents continuously collect and buffer performance metrics before power consumption calculation is needed. Metrics such as CPU usage, memory access, and I/O operations are accumulated in real-time and stored in data structures ready for processing. When power consumption attribution is required, the system uses pre-collected metrics combined with infrastructure power data to rapidly calculate request-level power consumption, significantly reducing processing time compared to collecting all necessary data at the moment of calculation

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250117551A1Machine learning model-based determination of power consumption in addressing application request(s)

Publication Date: 2025.04.10 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A machine learning power consumption model and process are provided for determining power consumption of computing environment resources in handling requests of an application. The process includes training the machine learning power consumption model to estimate power consumption of computing environment resources in handling requests. The training uses a training dataset derived from historical request-related data, and the training dataset includes request-related data and resource power consumption data. In addition, the process includes analyzing the requests. The analyzing includes, for a particular request of the requests, obtaining a common pattern of resource use and collecting real-time trace metrics to facilitate allocating resource use to the particular request. In addition, the process includes using the machine learning power consumption model, the obtained common pattern, and the collected real-time trace metrics for the particular request in generating an estimate of the power consumption in addressing the particular request.