Datacenter Power Management Using Infrastructure-Aware Current Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional power management systems in datacenters result in stranded power due to oversizing of server devices, inadequate power grid redundancy, and lack of accurate input current limit determination, leading to server device unavailability during power infrastructure issues.
Innovation Solution
A power management system that identifies power requirements and infrastructure limitations to set precise input current limits for each computing device, ensuring optimal performance and redundancy, with a hardware backup subsystem for rapid response during power failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If circuit breakers are sized for failover based on total PSU capacity, then failover capability is improved, but stranded power increases due to oversized PSUs
Solution Approach 1:
The patent changes the parameter used for circuit breaker sizing from total PSU capacity to actual input current consumption. By monitoring and using real current draw data rather than rated capacity, the system enables circuit breakers to be sized based on actual load requirements, eliminating stranded power while maintaining failover capability.
Solution Approach 2:
The patent replaces conventional mechanical/current-based circuit breaker sizing with a software-based power management system that monitors actual power consumption. This substitution allows dynamic adjustment of power limits based on real-time data, enabling precise circuit breaker sizing that prevents energy waste.
2Loss of energy
If server device power is capped to reduce stranded power, then energy efficiency is improved, but server device availability deteriorates during power infrastructure issues
Solution Approach 1:
The patent implements a feedback mechanism where the power management system continuously monitors actual input current consumption and adjusts power limits accordingly. This feedback loop enables the system to distinguish between normal operational current and fault conditions, allowing power capping to reduce stranded power while maintaining availability during infrastructure issues through intelligent, data-driven decisions.
3Device complexity
If single per-server-device power limit levels are used, then system simplicity is improved, but power grid fault tolerance deteriorates
Solution Approach 1:
The patent applies local quality by assigning individualized power limits to each server device based on its actual measured input current consumption rather than using uniform limits. This enables precise power management tailored to each device's specific characteristics and power grid connection, improving fault tolerance while maintaining manageable system complexity through automated profiling.
Data Source
AI summary
A power management system includes a power infrastructure having power infrastructure components that couple computing devices to power source(s), and a power management subsystem coupled to the computing devices. The power management subsystem identifies power requirement information for each of the computing devices and power transmission limitation information for the power infrastructure components, and determines a power infrastructure architecture that identifies how the power infrastructure components couple the computing devices to the power source(s). The power management subsystem then uses the power requirement information, the power transmission limitation information, and the power infrastructure architecture to generate and provide a respective input current limit to each of the computing devices. Each respective input current limit is configured to maximize performance of the respective computing device for which it was generated while ensuring availability of the power infrastructure in the event of an unavailability of a subset of the power infrastructure components.