Hierarchical Power Management for Datacenter Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed computing systems, such as those for cloud computing and high-performance computing, face challenges in managing power consumption efficiently, particularly in maintaining power within a narrow band to avoid high demand charges and fluctuations, which conventional solutions like capacitors and UPS systems are inadequate to handle, especially for large datacenters.
Innovation Solution
Implementing a hierarchical power management system that includes a facility power manager, system power-performance manager, and job power-performance manager to control power consumption within a specified band and rate of change using a combination of energy storage and generation, with green and non-green activities such as energy waste and local energy generation/storage to maintain power levels, and employing a power balloon module to adjust power consumption at facility, system, and node levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional solutions like capacitors and UPS systems are used to manage power consumption, then power stability is improved, but device complexity and cost increase significantly
Solution Approach 1:
The patent replaces traditional mechanical/electrical power stabilization systems (capacitors, UPS) with a software-based hierarchical power management system that uses monitoring and control algorithms to manage power consumption dynamically, thereby reducing hardware complexity while maintaining power stability
Solution Approach 2:
The power management system enables the data center to self-regulate its power consumption through automated monitoring and control mechanisms that adjust workload distribution and power allocation without requiring external intervention or extensive infrastructure
2Device complexity
If power management infrastructure is reduced to avoid expensive equipment, then device complexity is reduced, but power control precision deteriorates
Solution Approach 1:
The patent divides the power management system into hierarchical levels (facility-level, system-level, and component-level managers) that collectively provide precise power control without requiring a single complex infrastructure, thereby maintaining control precision while reducing overall system complexity
Solution Approach 2:
The system implements continuous monitoring and feedback mechanisms that track power consumption in real-time and automatically adjust power allocation and workload distribution, ensuring precise power control without extensive hardware infrastructure
3Loss of energy
If power consumption is strictly controlled within narrow bands, then energy efficiency is improved, but productivity decreases due to job execution constraints
Solution Approach 1:
The patent implements dynamic power management that adapts power band constraints in real-time based on workload characteristics, job priorities, and system state, allowing the system to maintain energy efficiency while flexibly accommodating productivity requirements when necessary
Solution Approach 2:
The system changes power management parameters (such as power band width, allocation ratios, and control thresholds) based on operational conditions, enabling the balance between energy efficiency and productivity to be optimized dynamically rather than fixed
4Manufacturing precision
If hierarchical power management system is implemented, then power control precision is improved, but device complexity increases
Solution Approach 1:
The hierarchical power management system is segmented into distinct functional levels (facility-level, system-level, component-level) with clearly defined responsibilities, allowing precise power control to be achieved through coordinated simple actions at each level rather than a single complex control mechanism
Data Source
AI summary
A system and method for computing at a facility having systems of multiple compute nodes to execute jobs of computing. Power consumption of the facility is managed to within a power band. The power consumption may be adjusted by implementing (e.g., by a power balloon) activities having little or no computational output.