Power Capping Feedback Normalization for Server Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power capping systems face challenges in adjusting the gain of closed-loop feedback control, leading to slow responses or oscillations, as a result of either under or over-gain settings, which complicates the process of efficiently limiting power consumption in servers.
Innovation Solution
A power capping system that includes a power controller and a management interface, utilizing a normalization factor and a predetermined gain constant to normalize error calculations, allowing for the generation of a power capping signal through pulse-width modulation, thereby setting the control percentage of the power capping signal based on the normalized error, and includes a calibration procedure to determine the normalization factor by measuring power consumption at maximum and minimum levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the gain of closed-loop feedback control is increased to improve response speed, then the controller's response becomes faster, but oscillation occurs that may not settle
Solution Approach 1:
The system dynamically adjusts the gain parameter based on operating conditions. The normalization factor is calculated from actual power consumption measurements and used to scale the error signal, effectively adapting the controller gain to the current state of the server and power supply conditions, preventing oscillation while maintaining responsive control
2Stability of the object's composition
If the gain of closed-loop feedback control is decreased to prevent oscillation, then stability is improved, but the controller's response becomes too slow
Solution Approach 1:
Rather than using a fixed low gain, the system uses a dynamic normalization factor that scales with actual power consumption. This allows the controller to respond quickly when power consumption is high (large normalization factor) while maintaining stability when power consumption is low (small normalization factor), effectively resolving the speed-stability tradeoff
3Measurement precision
If individual tuning of gain parameters is performed for each server to optimize performance, then control precision is improved, but device complexity and tuning time increase
Solution Approach 1:
The system performs self-tuning by automatically calculating the normalization factor from its own power consumption measurements during normal operation. The calibration procedure measures actual power consumption at different control percentages and derives the normalization factor without requiring external intervention or manual adjustment, enabling each server to optimize its own control parameters
Solution Approach 2:
The system changes the gain parameter dynamically based on measured power consumption characteristics. By using the normalization factor derived from actual measurements, the controller adapts to the specific electrical and thermal characteristics of each server, achieving precise control without manual tuning
Data Source
AI summary
A power capping system (10) and method (200) are provided. In one embodiment, a power capping system (10) includes a power controller (16) configured to calculate an error between a predefined maximum desired power and a power feedback signal associated with actual power consumption of a server (12) and to provide a power capping signal that substantially limits the power consumption of the server (12) based on a predetermined gain constant and the error. The system also includes a management interface (18) configured to generate a normalization factor based on the power feedback signal. The normalization factor can be implemented to normalize the error.