Phase-Aware Computing Scheduling for Energy and Thermal Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face significant energy inefficiencies due to overprovisioning and inefficient cooling, with existing solutions being primarily reactive and failing to account for thermal lags and varying cooling needs across servers and racks.
Innovation Solution
A holistic approach that treats computing and cooling energy as a first-class resource, using predictive thermal models and global schedulers to allocate energy budgets, and proactive/reactive control mechanisms to optimize energy efficiency by dynamically matching cooling efforts to thermal conditions and workload.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If servers are overprovisioned with high capacity and peak performance design, then peak capacity and performance are improved, but energy efficiency deteriorates due to gross energy inefficiencies at non-peak loading levels
Solution Approach 1:
The patent applies dynamic voltage and frequency scaling (DVFS) to make server components dynamically adjustable rather than static. The system continuously monitors workload demands and adjusts processor voltage and frequency in real-time, allowing servers to operate at optimal performance levels matching actual needs rather than fixed peak capacity designs. This dynamic adaptation resolves the contradiction by enabling high peak performance when needed while maintaining energy efficiency during lower utilization periods.
Solution Approach 2:
The patent changes operating parameters of server components, specifically voltage and frequency settings, based on workload conditions. By implementing multiple performance states and dynamically transitioning between them, the system can adjust energy consumption to match actual computational demands. This parameter adjustment allows the same hardware to deliver peak performance when required while operating efficiently at lower loads, resolving the overprovisioning energy waste problem.
2Temperature
If traditional reactive cooling solutions are used that adjust based on sensor feedback, then cooling responses are provided, but thermal lags are not accounted for and energy efficiency deteriorates
Solution Approach 1:
The patent implements predictive cooling control that uses thermal models to anticipate future thermal conditions before they actually occur. Instead of waiting for temperature sensors to detect overheating and then reacting, the system predicts thermal trends based on workload patterns and proactively adjusts cooling. This preliminary action accounts for thermal lags by preparing cooling capacity in advance, resolving the contradiction between providing adequate cooling response and minimizing energy consumption.
Solution Approach 2:
The patent employs advanced feedback mechanisms that combine real-time temperature sensor data with predictive thermal modeling. The system continuously monitors actual thermal conditions and compares them with predicted values, using this feedback to refine cooling control decisions. This dual feedback approach ensures adequate cooling response while optimizing energy usage by avoiding both over-cooling and under-cooling scenarios that plague traditional reactive systems.
3Area of stationary object
If servers are packed into smaller form factors to increase density, then space utilization is improved, but power dissipation worsens due to increased server count in given physical space
Solution Approach 1:
The patent applies segmentation by dividing the data center into manageable zones or racks, each with its own independent power and cooling management. This allows targeted optimization where high-density server segments can be closely packed with efficient cooling, while lower-density segments use less aggressive configurations. The segmented approach resolves the contradiction by enabling high space utilization through dense packing while controlling overall power dissipation through localized management and selective high-density deployment.
4Loss of energy
If DC power sources are used for servers, then energy dissipation is reduced, but the overall energy efficiency improvement is limited without coordinated cooling management
Solution Approach 1:
The patent merges power management and cooling management into a unified, coordinated control system. Instead of treating DC power conversion and cooling as separate optimizations, the system integrates them so that power delivery and cooling provision work together synergistically. This merging resolves the contradiction by ensuring that reductions in energy dissipation through DC power are fully realized in overall energy efficiency through coordinated cooling that matches actual thermal loads rather than operating independently.
Data Source
AI summary
A system, and method for controlling a computing system, comprising: reading a stored energy-performance characteristic of a plurality of different phases of execution of software, an execution of each phase being associated with a consumption of a variable amount of energy in dependence on at least a processing system performance state, the performance state being defined by a selectable performance-energy consumption optimization for at least two processing system components; scheduling a plurality of phases of execution of the software, in dependence on the stored energy-performance characteristics, for each of the respective phases of execution of the software and at least one system-level energy criterion; and executing the phases of execution of the software in accordance with the scheduling.