Inferential Power Monitoring via Instrumentation Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring power consumption in datacenters are either expensive or unable to accurately track power usage at the individual component level, leading to over-provisioning of cooling systems due to conservative estimates and lack of voltage and current sensors in all servers.
Innovation Solution
A system that infers power consumption from instrumentation signals using a nonlinear, nonparametric regression technique and an inferential power model generated during a training phase, pre-processing signals with an analytical re-sampling program, and utilizing software variables such as load metrics and hardware variables like temperature and voltage to estimate power consumption without the need for hardware power monitors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If hardware power monitors are used to measure power consumption, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces the mechanical/electrical hardware power monitor with a software-based inferential system. The system uses existing instrumentation signals (CPU utilization, memory usage, disk activity, network traffic) and applies machine learning models to estimate power consumption, eliminating the need for physical power monitoring hardware while achieving comparable or superior measurement precision.
Solution Approach 2:
The patent introduces an intermediary inferential system that mediates between existing instrumentation signals and power consumption measurement. Instead of directly measuring power with hardware, the system uses software intermediaries (regression models, neural networks) that process available system metrics to infer power consumption, thereby avoiding the need for direct electrical measurement hardware.
2Measurement precision
If voltage and current sensors are installed in power supplies, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent enables the system to self-measure power consumption using its own existing instrumentation infrastructure. By leveraging signals already present in the system (CPU metrics, memory status, disk activity, network I/O), the system eliminates the need for external sensors or additional measurement hardware, achieving self-sufficient power monitoring.
Solution Approach 2:
The patent makes existing instrumentation signals serve multiple functions. The same signals used for system performance monitoring and management are simultaneously utilized for power consumption estimation, eliminating the need for dedicated power measurement sensors and maximizing the utility of existing system infrastructure.
3Reliability
If cooling capacity is over-provisioned to meet maximum power ratings, then reliability is improved, but energy efficiency deteriorates
Solution Approach 1:
The patent transitions from static cooling capacity provisioning (based on maximum theoretical power ratings) to dynamic cooling capacity adjustment (based on real-time actual power consumption). By continuously monitoring system metrics and estimating actual power usage, the system enables cooling systems to dynamically adapt their capacity, providing sufficient cooling when needed while reducing cooling energy consumption when full capacity is not required.
Solution Approach 2:
The patent changes the parameter used for cooling capacity determination from conservative maximum power ratings to actual measured power consumption estimates. This parameter change enables more accurate matching of cooling capacity to actual thermal load, preventing both over-provisioning and under-provisioning of cooling resources.
4Ease of operation
If faceplate power ratings are used for cooling provisioning, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where actual system performance metrics are continuously monitored and fed into inferential models to estimate actual power consumption. This feedback loop replaces static faceplate ratings with dynamic, data-driven power estimates, maintaining ease of operation through automated software-based measurement while dramatically improving measurement precision.
Data Source
AI summary
A system that facilitates estimating power consumption in a computer system by inferring the power consumption from instrumentation signals. During operation, the system monitors instrumentation signals within the computer system, wherein the instrumentation signals do not include corresponding current and voltage signals that can be used to directly compute power consumption. The system then estimates the power consumption for the computer system by inferring the power consumption from the instrumentation signals and from an inferential power model generated during a training phase.