Power Utilization Index for Server Energy Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges in estimating energy consumption by servers for given workloads, as existing metrics are not normalized, making it difficult to compare and manage energy usage across diverse server configurations and manufacturers.
Innovation Solution
A power utilization index, referred to as Portable Energy Consumption (PEC), is introduced, which defines the energy consumption per unit of workload, allowing for the prediction of future power usage based on projected workload demands and enabling efficient workload assignment across servers from different vendors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing energy consumption metrics are used for servers, then energy usage can be measured, but the metrics are not normalized making it difficult to compare and manage energy usage across diverse server configurations and manufacturers
Solution Approach 1:
The patent transforms raw energy consumption measurements into a normalized Power Utilization Index by changing the parameter representation. The PUI is calculated by dividing power consumption by a workload metric, creating a standardized parameter that enables comparison across different server configurations, manufacturers, and generations while preserving the underlying energy measurement accuracy.
2Productivity
If server volumes grow to host more computers in data centers, then hosting capacity increases, but overall power consumption continues to grow
Solution Approach 1:
The system implements feedback by continuously monitoring power consumption and workload metrics, calculating the Power Utilization Index, and using this information to predict future power usage. This feedback loop enables data center operators to make informed decisions about server provisioning, workload allocation, and capacity planning to optimize the balance between hosting capacity and power consumption.
Solution Approach 2:
The patent applies preliminary action by predicting future power usage based on the Power Utilization Index and projected workload demands before actually deploying additional servers or increasing capacity. This allows operators to plan and prepare for power consumption increases in advance, potentially taking preventive actions such as optimizing existing server utilization or pre-provisioning power infrastructure.
3Adaptability or versatility
If servers from different manufacturers, generations, and configurations are deployed, then versatility and hosting capability improve, but energy consumption varies making management difficult
Solution Approach 1:
The Power Utilization Index serves as a universal metric that can be applied across servers from different manufacturers, generations, and configurations. By normalizing energy consumption data through the PUI calculation methodology, the system creates a multi-functional measurement tool that works universally across diverse server environments, enabling consistent energy management practices regardless of the specific server hardware being used.
Data Source
AI summary
A power utilization index is determined for a server that defines an amount of energy that is consumed by the server for a unit of workload performed by the server. Future power usage by the server may then be predicted based on the power utilization index and a projected workload demand on the server. Moreover, workload for the server may be selectively assigned in response to the predicting. The power utilization index may be determined by a obtaining measurements of power consumed by the server in response to various workloads. Measurements of workload demands placed on the server are also obtained for the workloads. The measurements of workload demands may separately account for demands placed upon a processor subsystem of the server, a memory subsystem of the server, a network communication subsystem of the server and a storage subsystem of the server, for the various workloads.