Data Center Server Provisioning Using AI Power Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data center management systems fail to adequately address the specific servers' impact on power consumption, making it difficult to optimize energy efficiency.
Innovation Solution
A machine learning model is trained and fine-tuned using temperature and power consumption data to predict the most energy-efficient server operations, enabling intelligent selection and shutdown of servers based on predicted power consumption differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional server selection methods (best-fit, worst-fit, round-robin) are used, then server resource allocation is achieved, but power consumption optimization is insufficient
Solution Approach 1:
The patent replaces conventional mechanical/server-based selection methods (best-fit, worst-fit, round-robin) with an AI-based machine learning model that predicts power consumption. This substitution enables the system to optimize power consumption by selecting servers based on predicted energy efficiency rather than simple allocation rules, directly addressing the insufficient power consumption optimization in conventional methods.
Solution Approach 2:
The patent introduces an AI model as an intermediary between server resource requests and server selection. This intermediary processes temperature information, server positions, and power consumption data to predict the impact of powering on/off servers on overall data center power consumption, enabling optimized decision-making without directly modifying the underlying server infrastructure.
2Use of energy by moving object
If servers are powered ON/OFF based on simple demand, then resource availability is maintained, but overall data center power consumption is not optimized
Solution Approach 1:
The patent employs preliminary action by training the machine learning model in advance using historical data from multiple data centers. This pre-trained model can then quickly predict power consumption impacts when making real-time server selection decisions, enabling the system to optimize power consumption without requiring complex real-time calculations that would compromise decision speed.
Solution Approach 2:
The patent implements feedback by continuously collecting temperature information, server positions, and power consumption data from the data center environment. This feedback loop allows the AI model to learn from actual outcomes and improve its predictions of how powering on or off specific servers affects overall data center power consumption, progressively enhancing measurement precision.
Data Source
AI summary
A data center has a management system for selecting a server computer to start or shut down. The management system has a machine learning model that is trained to predict power consumption of the data center using a large training dataset of many, different data centers. The machine learning model is fine-tuned using data of server computers of the data center. Input data that include temperature information of a server computer and position of the server computer are input to the machine learning model to obtain a predicted difference in power consumption of the data center. Predicted differences in power consumption of the data center are compared to select a server computer to start or shut down.