Priority-Aware Power Capping for Hierarchical Data Center Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current power capping solutions in data centers fail to efficiently allocate power to devices due to their hierarchical power distribution networks, leading to suboptimal utilization and increased infrastructure costs, as they do not consider the priority and characteristics of individual power consumption devices.
Innovation Solution
Implementing a priority-aware power capping mechanism that allocates power based on the priority and characteristics of each device within the hierarchical power distribution network, allowing for dynamic adjustment of power budgets to adhere to design limitations and contractual limits, utilizing a combination of hardware and software configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional power capping solutions are used, then infrastructure costs are reduced, but power allocation efficiency deteriorates due to failure to consider hierarchical power distribution network characteristics and device priorities
Solution Approach 1:
The power capping mechanism is segmented into multiple priority levels (e.g., priority 0 through priority 7), where each level corresponds to specific power consumption devices. This segmentation allows the system to allocate power budgets differently based on device priority, thereby improving power allocation efficiency without requiring complete redesign of the power distribution infrastructure.
Solution Approach 2:
The patent implements local quality by assigning different power budget allocation rules to different priority levels and device types. High-priority devices receive guaranteed minimum power budgets and can access remaining power budgets, while low-priority devices receive power budgets only when power is available. This localized differentiation resolves the contradiction by making the power capping mechanism adaptive to specific device requirements rather than applying uniform rules.
2Productivity
If power is allocated without considering device priority, then infrastructure costs are reduced, but power performance deteriorates due to suboptimal utilization
Solution Approach 1:
The power budget allocation is made dynamic through the priority-aware mechanism. Power budgets are not statically assigned but dynamically adjusted based on device priority levels and actual power availability. High-priority devices can dynamically access remaining power budgets when needed, while low-priority devices dynamically receive power when surplus is available. This dynamic approach improves power performance by ensuring critical devices always have adequate power while minimizing energy waste through efficient utilization of available power budgets.
Solution Approach 2:
The system implements feedback mechanisms where power consumption data from devices at different priority levels is continuously monitored. This feedback enables the power capping mechanism to adjust power budget allocations in real-time, ensuring that power is allocated efficiently according to actual device needs and priority levels, thereby improving power performance and reducing unutilized power budgets.
3Reliability
If minimum power budget is assigned to all devices, then device operation reliability is improved, but overall power utilization deteriorates due to power budget constraints
Solution Approach 1:
The patent applies parameter changes by modifying the power budget allocation parameters based on device priority levels. Instead of using a single fixed power budget parameter for all devices, the system uses multiple parameters (minimum power budgets, remaining power budgets, priority levels) that can be adjusted independently. This allows the system to maintain reliable operation for high-priority devices with guaranteed minimum power budgets while improving overall power utilization by allocating remaining power budgets to additional devices based on their priority and power availability.
Data Source
AI summary
A mechanism is provided for throttling power utilized by a set of power consumption devices using priority-aware power capping. Responsive to unassigned power budget remaining in the overall power budget after a minimum power budget value has been assigned to the child device based on an associated priority of the child device, an additional power budget value equal to a remaining priority-based exposed power demand value of the child device is assigned to the child device in response to the remaining unassigned power budget being greater than or equal to the remaining priority-based demanded power value thereby forming a total power budget for the child device. Responsive to design limitations of power distribution equipment in the data processing system or contractual limits of the data processing system being reached, a throttling is implemented by each child device based on the total power budget assigned to the child device.