Server Fan Weight Optimization Using CMA-ES for Hotspot Cooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for managing fan weights in servers lead to unnecessary energy consumption due to improper weight settings and complex configurations, often reverting to synchronous control, which fails to differentiate cooling demands across different hotspots.
Innovation Solution
A system and method using evolutionary strategy to generate and optimize fan weight combinations through random generation, testing, and selection using a covariance matrix adaptation evolution strategy (CMA-ES) algorithm to find the most energy-efficient settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If synchronous control is used to manage fan speeds, then control simplicity is maintained, but fan energy consumption increases due to inability to differentiate cooling requirements
Solution Approach 1:
The patent applies local quality by assigning different weight values to different fans based on their spatial relationship with hotspots. Each fan receives a customized weight parameter that reflects its local cooling importance, enabling differentiated control rather than uniform synchronous control. This resolves the contradiction by maintaining simple weight-based control logic while achieving energy-efficient differentiated fan speed management.
Solution Approach 2:
The patent changes the control parameter from uniform fan speed to weighted fan speed, where each fan's speed is adjusted according to its assigned weight parameter. The evolutionary algorithm optimizes these weight parameters to minimize energy consumption while maintaining thermal safety, thus resolving the contradiction between control simplicity and energy efficiency.
2Use of energy by moving object
If Intel's fan weighting method is used to differentiate cooling requirements, then energy consumption can be reduced, but configuration complexity increases due to output weight matrix setup
Solution Approach 1:
The patent applies self-service by using an evolutionary algorithm that automatically generates and optimizes fan weight configurations without requiring manual setup of weight matrices. The system self-configures the optimal weights based on thermal safety constraints and energy consumption objectives, eliminating the complex manual configuration process while achieving energy efficiency.
Solution Approach 2:
The patent performs preliminary action by pre-calculating optimal fan weights through evolutionary optimization before actual fan control operation. The system prepares the weight configuration in advance, storing it for subsequent use, thus avoiding complex real-time calculations and simplifying the operational configuration process.
3Ease of operation
If manual experience-based weight settings are used, then configuration simplicity is maintained, but fan control performance deteriorates due to suboptimal weight values
Solution Approach 1:
The patent applies feedback by using thermal safety constraints and energy consumption metrics to guide the evolutionary optimization of fan weights. The system continuously evaluates weight configurations against thermal and energy objectives, using this feedback to iteratively improve weight settings. This resolves the contradiction by maintaining simple weight-based configuration while achieving superior control performance through data-driven optimization.
4Reliability
If the number of temperature points and fans is increased, then cooling coverage is improved, but weight matrix configuration becomes more complex
Solution Approach 1:
The patent applies self-service by using the evolutionary algorithm to automatically handle the configuration of weights for any number of fans and temperature points. The system self-adapts to the specific hardware configuration without requiring manual setup of complex weight matrices, thus maintaining cooling coverage while eliminating configuration complexity regardless of system scale.
Data Source
AI summary
A method of setting fan weights based on evolutionary strategy, performed by a computing element, includes: generating first weight combinations randomly, each of the first weight combinations including weights configured to set operation parameters of fans, conducting a fan control test using the first weight combinations and receiving test results, calculating fitness values with a fitness function according to the first weight combinations and the test results, selecting candidate weight combinations from the first weight combinations, with the candidate weight combinations corresponding to lowest fitness values among the fitness values, generating second weight combinations according to the candidate weight combinations using a covariance matrix adaptation evolution strategy (CMA-ES) algorithm, and using one of the second weight combinations to control the fans when a generation number corresponding to the second weight combinations reaches a threshold value.