Dynamic Power Cap Allocation for Computing Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power management systems for computing systems fail to efficiently allocate power based on varying power requirements and changes in the system, leading to potential overcurrent events and operational inefficiencies, especially when different types and models of servers are updated or replaced.
Innovation Solution
A power management method and system that determines a priority sequence for computing nodes based on their minimum power consumption and allocates power cap values dynamically, ensuring that each node receives sufficient power without exceeding the total available power, by iteratively updating the total power and power ratio.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If power is allocated without considering varying power requirements of different computing nodes, then allocation simplicity is maintained, but power efficiency and reliability deteriorate due to overcurrent events and operational inefficiencies
Solution Approach 1:
The patent dynamically adjusts power allocation parameters (power cap values) based on varying system conditions, computing node types, and power requirements. The power management controller modifies allocation parameters in real-time to optimize power efficiency while adapting to different server models and update scenarios.
Solution Approach 2:
The power allocation system transitions from static to dynamic allocation, where power cap values are continuously adjusted based on current system state, computing node power requirements, and available power capacity. This dynamic approach prevents overcurrent events while maintaining operational efficiency.
2Reliability
If power cap values are allocated without priority sequencing, then allocation speed is maintained, but power distribution reliability deteriorates when nodes are added or removed
Solution Approach 1:
The patent pre-establishes priority sequences for computing nodes based on their power requirements and system importance before power allocation occurs. This preliminary sequencing enables rapid power cap assignment when nodes are added or removed, maintaining reliability without time-consuming real-time negotiations.
Solution Approach 2:
The power allocation process is segmented into discrete priority levels and sequential steps, allowing the system to systematically allocate power caps to computing nodes in a predetermined order. This segmentation ensures reliable power distribution while maintaining efficient allocation speed.
3Measurement precision
If total power is not iteratively updated during allocation, then calculation complexity is reduced, but power cap accuracy deteriorates leading to insufficient or excessive power allocation
Solution Approach 1:
The power management controller implements iterative feedback loops where total power availability is continuously updated as power caps are allocated to individual computing nodes. Each allocation decision feeds back into the total power calculation, ensuring subsequent allocations are based on accurate remaining power capacity, thereby achieving precise power cap distribution.
Solution Approach 2:
The patent performs multiple iterative calculations of total power and power ratios, allocating power caps in successive approximations rather than a single step. This partial action approach progressively refines power cap accuracy for each computing node while managing calculation complexity through structured iteration.
Data Source
AI summary
A power management system and method including: determining a maximum power function with respect to a computing node; determining a power cap value, wherein the power cap value is the greater of the maximum power function and a minimum power consumption value of the computing node; and allocating the power cap value to the computing node, wherein the maximum power function is a product of a total power and a power ratio of the computing node.