RAN Node Energy Accounting Across AI/ML Pipeline Stages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current energy monitoring techniques for Radio Access Networks (RANs) lack the capability to comprehensively log and track the total energy consumed by deployed AI/ML pipelines and their associated processes, failing to consider the ripple effect of energy consumption across connected entities and lacking standardized mechanisms for comprehensive energy accounting.
Innovation Solution
A method and apparatus for energy accounting in RANs that measure energy consumption at various stages of AI/ML pipelines, including data collection, training, inference, and execution, and across network nodes, enabling a holistic view of energy consumption and savings by computing a net energy measure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If AI/ML pipelines are deployed in RAN to optimize network operations, then network intelligence and automation are improved, but energy consumption increases due to computational overhead
Solution Approach 1:
The patent changes the parameter of energy accounting from local node-level measurement to network-wide cumulative measurement. By aggregating energy consumption across multiple network nodes and AI/ML pipeline stages, the system transforms the parameter perspective to reveal the true net energy impact, enabling informed decisions about whether AI/ML deployment is energetically justified
Solution Approach 2:
The patent implements a feedback mechanism where energy consumption data from AI/ML pipelines is collected, aggregated, and used to compute net energy metrics. This feedback loop allows the network operator to assess whether the energy savings from AI/ML-optimized network operations exceed the energy consumed by the AI/ML pipelines themselves, enabling continuous optimization
2Measurement precision
If energy monitoring is implemented at individual network nodes, then local energy efficiency is measured, but comprehensive energy accounting across the network is lost
Solution Approach 1:
The patent merges energy consumption data from multiple individual network nodes and AI/ML pipeline stages into a cumulative network-wide measurement. By combining local measurement precision with network-wide aggregation, the system preserves detailed energy information while establishing a holistic view that captures the total energy impact of AI/ML deployment across the entire network
3Measurement precision
If detailed energy tracking at each AI/ML pipeline stage is implemented, then energy efficiency analysis is improved, but system complexity increases
Solution Approach 1:
The patent segments the AI/ML pipeline into distinct stages (data collection, training, inference, etc.) and measures energy consumption at each segment. This segmentation enables precise energy efficiency analysis by identifying which specific stages consume the most energy, allowing targeted optimization while maintaining manageable system complexity through modular measurement
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Approaches for energy accounting of network nodes in a Radio Access Network (RAN), such as an Open RAN, are described. In an example, a first value corresponding to energy consumed in implementation and during operation of one of an Al and a ML pipeline at a first network node of a RAN may be obtained. Thereafter, an estimate of energy saving for the first network node may be determined by comparing energy usage before and after executing a decision based on an inference of one of an Al and ML model deployed at the first network node. Based on the first value and the estimate of energy saving for the first network node, a measure of net energy may be computed. In response to the computed measure, a pre-defined action may be executed.