Computer Infrastructure Power Scheduling for Low-Carbon Energy Use
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for electrical energy and thermal management in computing infrastructures, such as supercomputers, leads to high carbon footprints and inefficiencies in energy consumption, necessitating a method to optimize electrical consumption while minimizing reliance on carbon-based grids.
Innovation Solution
A method and system that utilize local renewable energy sources, thermal regulation, and intelligent task scheduling to minimize grid energy use, incorporating environmental data and task criticality to adjust power consumption and thermal management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If computing infrastructure performs more tasks using computing resources, then computing productivity increases, but electrical energy consumption increases
Solution Approach 1:
The patent implements dynamic task scheduling that adjusts computing resource allocation based on real-time electrical energy availability from local sources and grid conditions. The system dynamically determines which tasks to execute, pause, or defer based on current energy constraints, enabling productivity optimization without excessive energy consumption.
Solution Approach 2:
The system changes operational parameters by adjusting computing resource utilization levels based on energy availability. When local energy sources are sufficient, the system increases computing resource usage to maximize productivity. When energy availability decreases, the system reduces resource usage or pauses non-critical tasks to maintain operational sustainability.
2Productivity
If computing infrastructure performs more tasks, then computing productivity increases, but heat dissipation increases
Solution Approach 1:
The patent implements dynamic thermal management that adjusts task execution based on real-time temperature conditions. The system monitors heat dissipation levels and dynamically determines whether to continue, pause, or defer tasks based on thermal constraints, preventing overheating while maintaining maximum feasible productivity.
Solution Approach 2:
The system uses feedback from temperature sensors to adjust computing resource allocation and task scheduling. When temperature exceeds thresholds, the system receives feedback and automatically reduces resource usage or pauses tasks to allow cooling, creating a closed-loop thermal management system.
3Reliability
If thermal management system operates to maintain optimal temperature, then computing infrastructure reliability increases, but electrical energy consumption increases
Solution Approach 1:
The patent implements dynamic thermal management that adjusts cooling system operation based on real-time temperature conditions and energy availability. The system operates thermal management at minimum necessary levels during low-energy periods and increases cooling capacity when energy is abundant, maintaining reliability while minimizing energy consumption.
4Productivity
If computing infrastructure uses more electrical energy from the grid, then computing productivity increases, but carbon footprint increases
Solution Approach 1:
The patent implements preliminary action by capturing and storing electrical energy from local sources (such as on-site generation or energy storage systems) before it would otherwise be unavailable or more expensive. This stored energy is then used to power computing tasks, replacing grid electricity and reducing carbon footprint while maintaining productivity.
Solution Approach 2:
The system converts the limitation of local energy source capacity into a benefit by using it as a constraint that drives optimization of task scheduling and resource allocation. This constraint forces the system to be more efficient and selective about task execution, ultimately reducing overall energy consumption and carbon footprint while maintaining essential productivity.
Data Source
Figure 1~2
AI summary
The invention relates to a method for optimizing the electrical consumption of a computer infrastructure (CU) required to execute a set (W) of computing tasks using computing resources (RES), the computer infrastructure (CU) and the computing resources (RES) being powered by an electrical network (GRID) and at least one local electrical energy source (ENR), the method (100) comprising: - the determination (230) of a subset (W') of computing tasks to be scheduled on the plurality of computing resources (RES) as a function of electrical powers deliverable (PDGRID, PDENR) by the electrical network (GRID) and the local electrical energy source (ENR); and - the determination (290) of a parameterization (REP) of the electrical network (GRID) and the local electrical energy source (ENR) to power the computer infrastructure (CU).