Dynamic Power Cap Adjustment for Computing System Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing systems may shut down unexpectedly when the total power draw from components exceeds the capacity of the power subsystem, and existing solutions like aggressive throttling are inefficient.
Innovation Solution
A method and system that adjust power caps for computing system components by receiving an over-power signal from the power subsystem and reducing the power cap for high-power components by a predetermined amount to prevent excessive power draw.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If aggressive throttling is applied to prevent power overdraw, then power subsystem overload is prevented, but computing component performance is drastically reduced
Solution Approach 1:
The power cap for computing components is made dynamic rather than static. The BMC monitors power draw in real-time and adjusts the power cap dynamically - reducing it when power limits are approached and maintaining higher levels when power is available. This allows the system to prevent overdraw while avoiding unnecessary performance throttling.
Solution Approach 2:
The system implements a feedback loop where the BMC receives power draw information from the power subsystem, compares it against limits, and adjusts power caps accordingly. This closed-loop control enables precise power management that responds to actual system conditions rather than applying fixed aggressive throttling.
2Reliability
If power cap is reduced by a predetermined amount in response to over-power signal, then frequency of over-power events is reduced, but maximum power delivery capability is limited
Solution Approach 1:
The power cap adjustment is dynamic and conditional rather than a permanent reduction. When an over-power signal is received, the BMC reduces the power cap by a predetermined amount. However, when power draw returns to acceptable levels, the BMC can restore the power cap to higher levels, allowing the system to maintain reliability while preserving maximum power delivery capability when needed.
Solution Approach 2:
The predetermined reduction in power cap acts as a cushioning measure that prevents future over-power events. By proactively reducing the cap below the limit threshold, the system creates a safety margin that prevents exceeding power limits while allowing near-full power delivery during normal operation.
Data Source
AI summary
Systems and apparatus for adjusting power caps for computing system components includes: a power subsystem, one or more components, and a BMC (baseboard management controller) configured to: receive, from the power subsystem, an over-power signal indicating a draw on the power subsystem exceeding one or more limits of the power subsystem, and reduce, by a predetermined amount, a power cap for the one or more components.