Hybrid Rack Cooling Control for Pump-Fan Energy Minimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cooling systems in data centers are inefficient in reducing energy consumption while maintaining optimal temperature conditions for high-performance processors, as they rely on traditional air cooling and direct-to-chip liquid cooling, which are not sufficient to handle the heat generated by power-intensive processors.
Innovation Solution
An optimal controller is introduced to manage the hybrid liquid-air cooling system by optimizing the liquid pump speed and fan speeds of cooling fans, using a convex optimization model that minimizes total energy consumption while ensuring processor temperatures remain below a predetermined threshold, utilizing real-time data from computing nodes, cooling fans, and the coolant distribution unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional air cooling is used for high-power processors, then the cooling system is simple to implement, but the cooling performance is insufficient and energy consumption is high
Solution Approach 1:
The patent combines air cooling and liquid cooling systems into a hybrid cooling architecture. The liquid cooling subsystem directly contacts high-power processors through cold plates to remove concentrated heat, while air cooling handles ambient temperature control. This merging resolves the contradiction by achieving superior cooling performance through liquid cooling while maintaining system implementability through the established air cooling infrastructure.
Solution Approach 2:
The cooling system is segmented into distinct functional zones: liquid cooling is applied specifically to high-power processors generating concentrated heat, while air cooling serves general ambient temperature management. This segmentation allows each cooling method to be optimized for its specific function, resolving the contradiction between implementation simplicity and cooling performance sufficiency.
2Reliability
If direct-to-chip liquid cooling is implemented, then cooling performance improves and energy consumption decreases, but the system complexity increases
Solution Approach 1:
The patent implements dynamic control of the liquid cooling system with adjustable pump speeds and flow rates that adapt to real-time thermal conditions of processors. This dynamic operation allows the system to optimize cooling performance while reducing complexity by only activating liquid cooling when and where needed, rather than requiring always-on complex cooling infrastructure.
Solution Approach 2:
The patent introduces a thermal management controller as an intermediary that coordinates between air cooling and liquid cooling subsystems. This mediator optimizes the interaction between the two cooling methods, managing system complexity by centralizing control logic while allowing individual subsystems to remain relatively simple in their own operations.
3Ease of operation
If pump speed and fan speeds are not optimized, then the control system is simple to operate, but total energy consumption is high
Solution Approach 1:
The patent implements feedback control where thermal sensors continuously monitor processor temperatures and pump/fan speeds are automatically adjusted based on real-time thermal conditions. This feedback mechanism resolves the contradiction by enabling energy optimization through automated speed adjustment while maintaining ease of operation, as the system self-regulates without requiring manual intervention or complex user control.
Solution Approach 2:
The cooling system performs self-optimization of pump and fan speeds based on thermal feedback, eliminating the need for manual control or complex external management. The system serves itself by automatically adjusting operational parameters to minimize energy consumption while maintaining adequate cooling, thus resolving the contradiction between operational simplicity and energy efficiency.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively reduces the total energy consumption of the cooling system while maintaining processor temperatures within safe limits, optimizing both liquid and air cooling dynamics to achieve efficient heat removal in data centers.
Implementation Method 1
Direct-to-chip liquid cooling provides a better cooling performance
Implementation Method 2
a liquid flow rate is controlled via pump speed control in a coolant distribution unit (CDU)
Implementation Method 3
an airflow rate is controlled via fan speed control on the back of an electronic rack
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An optimal controller is utilized minimize a total energy consumption of a CDU and cooling fans by optimizing a liquid pump speed of the CDU and fan speeds of the cooling fans collectively, while satisfying a set of constraints associated with liquid pump and the cooling fans. The total energy consumption includes a liquid pump energy cost and a fan energy cost. The optimal controller is to control the pump speed and fan speed at the same time to minimize the total energy cost, while keeping the processor's temperature below a predetermined reference temperature using an optimization function or model. The optimization model includes both the liquid and air cooling dynamics and converted the dynamics to a convex optimization problem. The optimization is performed to determine a set of pump speed and fan speeds such that the objective function reaches minimum, while the constraints are satisfied.