CDU Pump Speed Control for Server Rack Cooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data center cooling systems face inefficiencies due to fixed pump speeds in coolant distribution units (CDUs), leading to wasted energy and inadequate cooling during varying workloads among IT components, particularly in high-power computing environments.
Innovation Solution
A novel liquid control architecture that adjusts the CDU pump speed based on the workload of IT components, using GPU card workload and temperature feedback to dynamically control liquid flow rates, ensuring efficient cooling and reducing energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a CDU pump operates at constant speed, then the liquid flow rate remains stable, but pumping energy is wasted and cooling efficiency decreases during low workload periods
Solution Approach 1:
The patent implements variable speed control for the CDU pump, allowing it to dynamically adjust its rotation speed based on real-time cooling demands. The pump controller receives temperature data from IT components and workload information, then modulates the pump speed accordingly. This dynamic operation enables the system to match liquid flow rate with actual cooling needs, eliminating energy waste during low workload while maintaining adequate cooling capacity during high demand periods.
2Loss of energy
If a CDU uses temperature difference feedback control, then partial pumping energy is saved, but the slow dynamic response and non-uniform liquid distribution fail to handle extreme workload conditions
Solution Approach 1:
The patent incorporates predictive cooling by receiving workload information in advance from the IT components. The pump controller uses this foresight to proactively adjust the pump speed before temperature criticality occurs, rather than merely reacting to temperature differences after they arise. This preliminary action enables the system to prepare adequate cooling capacity ahead of time, ensuring reliable cooling even during sudden extreme workload spikes that would overwhelm slower feedback systems.
Solution Approach 2:
The system implements a dual-feedback mechanism combining real-time temperature monitoring from IT components with workload information feedback. The pump controller continuously receives both temperature data and workload metrics, creating a comprehensive feedback loop that enables more responsive and accurate pump speed adjustment compared to temperature-difference-only control. This enhanced feedback allows the system to detect cooling demands earlier and respond more uniformly across all IT components.
3Temperature
If liquid cooling is implemented for high power density racks, then heat removal efficiency increases, but system complexity and energy consumption increase
Solution Approach 1:
The patent designs the CDU system to perform multiple functions: it provides both cooling and heating capabilities, manages variable pump speeds, monitors multiple temperature zones, and integrates with workload management systems. By creating a multi-functional cooling infrastructure, the system can handle diverse thermal management needs across different IT components and workload conditions, justifying the increased complexity through enhanced versatility and efficiency.
Solution Approach 2:
The system dynamically changes key operating parameters including pump rotation speed, liquid flow rate, and pump power consumption based on real-time conditions. By continuously adjusting these parameters to match actual cooling demands, the system optimizes heat removal efficiency while minimizing unnecessary energy consumption and operational complexity. This parameter adaptation allows the cooling system to scale its complexity only when and where thermal management demands require it.
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 solution provides flexible and adaptive cooling, reducing the risk of temperature overshoot and energy waste by aligning liquid flow rates with actual workload demands, thereby enhancing cooling efficiency and reducing operational costs in data centers.
Implementation Method 1
remove heat generated by information technology (IT) components using heat removal liquid
Implementation Method 2
Liquid heat removal offers a solution for higher power computing racks due to the relatively higher heat capacity and greater energy efficiency possible with liquid heat removal
Implementation Method 3
The CDU has the function of heat exchange between two loops via heat exchanger
Data Source
AI summary
A liquid control architecture is designed to control a liquid flow rate based on the workload of the IT components of an electronic rack, which provides sufficient cooling capability to each IT component, saving a total cost of pump energy. A coolant distribution unit (CDU) is included in each electronic rack to pump a liquid flow to a liquid supply manifold of the electronic rack. Each outlet on the supply manifold provides heat removal liquid to each IT component (e.g., one or more GPUs of a server blade) using flexible tubing. The liquid heat removal system utilizes the workload of each IT component to drive the CDU pump speed. In addition, the temperature of the IT components rather than return liquid temperature is utilized as a baseline to monitor the temperature of the electronic rack and change the control logic (e.g., liquid flow rate) as needed.