Dynamic Power Capping for Server Performance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In high-performance servers, accurately setting power priority and limitation ranges is challenging due to varying power consumption during high-load processing like machine learning, leading to potential performance deterioration.
Innovation Solution
A management program that calculates and adjusts power consumption estimates for servers, setting upper limits based on current processing states to prevent power capping that contradicts user intentions, thereby maintaining performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If power capping is set based on fixed priority order and limitation range, then power consumption is controlled within upper limit, but processing performance deteriorates during high-load operations
Solution Approach 1:
The patent implements dynamic power capping by continuously monitoring actual power consumption and adjusting the power limitation range in real-time based on current processing states, rather than using fixed priority orders and static limitation ranges. This allows the system to adapt power allocation to actual workload conditions, preventing performance deterioration during high-load operations while maintaining power control.
Solution Approach 2:
The system incorporates feedback mechanisms by monitoring actual power consumption values and using this information to dynamically adjust power capping settings. The management apparatus receives power consumption information from servers and uses this feedback to modify priority orders and limitation ranges, creating a closed-loop control system that balances power consumption with processing performance.
2Use of energy by stationary object
If administrator manually determines power priority and limitation ranges, then power allocation can be optimized, but system complexity and management overhead increase
Solution Approach 1:
The patent enables the system to automatically determine and adjust power priority orders and limitation ranges based on actual power consumption data and processing states, eliminating the need for administrators to manually configure these parameters. The management apparatus autonomously optimizes power allocation by analyzing server workload and consumption patterns, significantly reducing management overhead while maintaining efficient power distribution.
Solution Approach 2:
The system dynamically changes power management parameters (priority orders and limitation ranges) based on actual operating conditions rather than using fixed administrator-defined values. This automated parameter adjustment allows the system to optimize power allocation adaptively without requiring complex manual configuration or intervention.
3Ease of manufacture
If more high-performance servers are mounted in each rack, then installation cost is reduced, but power consumption increases beyond upper limit
Solution Approach 1:
The patent implements dynamic power capping that automatically adjusts power limitation ranges based on actual power consumption and server workload conditions. This allows data centers to maximize the number of high-performance servers in each rack while the system dynamically controls power allocation to stay within upper limits, enabling cost-effective server density increases without exceeding power capacity.
Data Source
AI summary
A non-transitory computer-readable recording medium stores a management program for causing a computer to execute a process the includes receiving designation of a first resource used by a specific server included in a plurality of servers, calculating a first estimated value of power to be consumed by the plurality of servers when the specific server executes first processing using the first resource, determining whether the first estimated value exceeds a first upper limit value of power to be consumed by the plurality of servers, setting in the specific server, when determined that the first estimated value does not exceed the first upper limit value, a second estimated value of power to be consumed by the specific server when the first processing is executed using the first resource, as a second upper limit value of power to be consumed by the specific server.