CPU Subset Clock Throttling for Data Center Power Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for reducing energy consumption in data centers by throttling CPUs operate in a binary manner, failing to balance CPU operations with actual operational demands, leading to inefficient energy usage.
Innovation Solution
A system that individually adjusts the clock speeds of CPUs based on operational demands, segregating them into subsets with different policies to maintain performance while minimizing energy consumption, using power control components and AI/ML to dynamically manage CPU operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional binary throttling methods are used to reduce energy consumption, then energy usage is reduced, but system performance and operational efficiency deteriorate due to failure to balance CPU operations with actual operational demands
Solution Approach 1:
The patent segments the CPU population into multiple subsets (e.g., first subset, second subset, third subset) with different operational policies. Each subset can be independently controlled with different clock speeds, allowing the system to reduce energy consumption by throttling specific subsets while maintaining high performance in others, thus resolving the contradiction between energy reduction and performance maintenance
Solution Approach 2:
The patent applies different operational qualities (clock speeds, policies) to different CPU subsets based on their specific operational demands. Rather than uniformly throttling all CPUs, the system tailors the operational state of each subset to its actual workload requirements, reducing energy consumption where possible while maintaining performance where needed
2Loss of energy
If CPU clock speeds are reduced to minimize energy consumption, then energy efficiency is improved, but the ability to satisfy operational demand deteriorates
Solution Approach 1:
The patent implements dynamic clock speed adjustment for different CPU subsets based on real-time operational demand assessment. The system continuously monitors demand data and dynamically reconfigures which subsets operate at high clock speeds versus low clock speeds, ensuring that energy efficiency is optimized without compromising the system's ability to satisfy operational demands
Solution Approach 2:
The system incorporates feedback mechanisms that monitor operational demand and system performance, using this information to adjust CPU subset configurations. This feedback loop ensures that energy consumption is minimized while maintaining the reliability needed to satisfy operational demands, as the system adapts its configuration based on actual performance requirements
3Reliability
If all CPUs are operated at high clock speeds to maintain system performance, then operational reliability is maintained, but energy consumption increases
Solution Approach 1:
The patent divides the CPU population into segments with different operational states, allowing the system to maintain high performance reliability through critical subsets operating at high clock speeds while reducing energy consumption through non-critical subsets operating at lower clock speeds
Solution Approach 2:
The system changes operational parameters (clock speeds, policies) of different CPU subsets based on assessed operational demands, transitioning between high-performance and energy-efficient states as needed to balance reliability and energy consumption
4Use of energy by moving object
If conventional binary throttling is applied uniformly to all CPUs, then energy consumption is reduced, but adaptability to different operational demands deteriorates
Solution Approach 1:
The patent segments CPUs into multiple subsets that can be independently configured with different policies and clock speeds, enabling the system to adapt to varying operational demands across different subsets while reducing overall energy consumption through selective throttling
Solution Approach 2:
The system applies local quality control by assigning different operational characteristics to different CPU subsets based on their specific operational contexts, thereby maintaining adaptability to diverse demands while achieving energy reduction through localized optimization
Data Source
AI summary
Various systems and methods are presented herein regarding controlling operation of central processing units (CPUs) to reduce power consumption at a data center(s). A first subset of CPUs located on a computer system can be operationally adjusted while a second subset of CPUs can be designated as having to be available at all times with a default operating condition (e.g., to run background operations). As operational demand placed on the computer system reduces, operation of the first subset of CPUs can be throttled back (e.g., clock speed reduced) while the second subset of CPUs remain at the default operating condition. As operational demand subsequently increases respective CPUs in the first subset of CPUs can have their operating condition (e.g., clock speed) increased. By adjusting the operating condition of one or more CPUs, power consumption at the data center can be reduced during periods of low operational demand.