Energy Efficient Thermal Management Of Data Centers – Data center optimization has always played an important role at headquarters. By optimizing the environmental controls of our data centers, we can reduce our environmental impact, while ensuring that people can always trust our products.
In most other complex systems, optimizing energy consumption is a process of trial and error. But experimenting with every component of a live data center is risky, as a single miscalculation can cause problems like power loss events. To ensure that our data centers are always performing at their best, we have developed a digital simulator that replicates the cooling and thermal behavior of the facilities. With our simulation, we can test control policies using AI or model predictive control – a set of algorithms that interact with the model to predict the response of the data center to certain conditions. Unlike similar systems trained on historical data, our simulator accurately represents even the most extreme circumstances, including hypothetical scenarios, and can make quantitative predictions for data centers that have yet to be built.
- 1 Energy Efficient Thermal Management Of Data Centers
- 2 Pdf) Thermal Metrics For Data Centers: A Critical Review
Energy Efficient Thermal Management Of Data Centers
Model predictive control has proven to be a powerful optimization strategy in recent years, but it requires a highly accurate model of the system to be effective. For the best possible simulation, we maximized the strengths of two different approaches. First, we developed a physics-based model—embedding it with equations describing thermal processes relevant to our data centers—so that our simulator would respect the principles of physics when representing new situations. We have combined this method with statistical data science to leverage the knowledge we have gathered in our large fleet of data centers.
Energy Efficient Thermal Management Of Air Cooled Data Centers
Using measurements from the many sensors in our facilities, we can create numerical models that predict the energy consumption of the building. These models, while highly accurate, may not generalize well to situations outside of the normal operating conditions reflected in the training data.
To quantify and visualize the fluid dynamics of our data centers, Meta works extensively with detailed physics-based models called Computational Fluid Dynamics (CFD) simulations. We can study previously unseen conditions and designs with these models, but they require a significant amount of computing power, and they take a long time to run.
Thermal simulators, or gray box models, open up a realm of new opportunities beyond data science and CFD approaches. They solve thermal balance equations to describe the conditions in a single room or even a row in a facility, producing results quickly and with little input data. We can use the output to understand the responses of our data centers to extreme weather conditions or to train a reinforcement learning algorithm, an AI system that finds the optimal solution through trial and error in a virtual environment.
Our dynamic model combines first-principle physics with modeling languages, including Modelica. To simulate a particular data hall, we first need static, site-specific details about the facility’s geometry, construction materials, and HVAC, as well as its system configurations and component efficiencies. We then numerically recreate the control strategies (using the control description language) that govern the behavior of all HVAC and water equipment as functions of indoor and outdoor conditions.
Why Invest In The Data Center Economy
Once the tasks are completed, we can start entering data. We choose variables whose influence we want to understand, such as temperature, energy and water consumption, and describe them as a time series.
Our data science-based models are generally very accurate when predicting the thermal response of a data hall to typical temperature ranges. But to keep our services running and prevent damage to hardware, we also need to understand how data centers behave in extreme conditions that put water and power infrastructure under high stress. We do not have the necessary data to train a data science-based model to represent severe cases, but a thermal simulator can simulate these situations and predict the effects of our response strategies.
To demonstrate the capability of our simulators, we studied the response of one of our data centers to a severe winter storm in Texas in February 2021. This unprecedented multi-day winter storm began with a temperature swing of 70 degrees Celsius, dropping the temperature below freezing for 10 days and reaching up to 2 ˚F.
We input the conditions recorded at midnight on February 7, 2021, and then let our thermal simulator run the next 15 days temperature setpoint time series, supply fan airflow, setpoint, and server load of the incoming supply airflow to data hall. We did not add operational data other than these four parameters.
Dupont And Jetcool Announce Collaboration In Thermal Management
Below is the command of our HVAC Economist – which controls the mixing of hot and cold air – trying to stabilize the internal temperature of the data hall. The signal varies between 15 and 100 percent, or nearly the full system range. The simulator was able to predict this behavior accurately.
It is important to closely monitor the temperature of the supply air entering the data halls, which is affected by both the outside temperature and the heat from the server rooms. To simulate this variable, we need to build a model of the complete control loop (the command above is an example), mechanical equipment and server room, using only physical principles, without fitting sensor data.
The figure below shows the simulated and measured supply air temperatures relative to the temperature setpoint during the 15 days we studied. Specify a mean absolute error (MAE) of 0.5˚F for the supply air temperature over the entire period. The MAE has a normal distribution centered at 0˚F (right panel), indicating no bias or systematic drift in our model for the 15-day forecast model period. In addition, our model accurately simulated the dynamics of the control system that regulates the temperature around the set point, as is particularly noticeable in the first and last days of this time period. Again, no operational data other than outdoor air conditions, temperature setpoint, supply fan airflow setpoint, and server load were entered into this simulation, demonstrating that the model was truly predictive in this extreme scenario.
The air temperature in the supply was stable compared to the parachute temperatures outside, and our simulator picked this up accurately as well.
Incooling Is Building Servers That Uses Phase Change To Cool Down
We also validated the simulator on 24 random dates during 2021. Our simulator accurately modeled the data center response on these dates—representing a wide range of weather conditions—and achieved an MAE of 1.3˚F for the supply air temperature.
According to the International Energy Agency, data centers consume about 1% of global electricity demand, which contributes to 0.3% of all global CO2 emissions. In this regard, Meta’s operational data centers have achieved net zero carbon emissions, are LEED Gold certified, and are powered by 100 percent renewable energy. They use 32% less energy and are 80% more water efficient on average than the industry standard. Meta is committed to recovering more water than we consume by 2030, and we hope our simulator will help us continue to reduce our resource consumption and ensure our operations are sustainable.
To further save resources, the physical modeling team at Meta is now training our simulator to predict and optimize the energy and water consumption of our data centers, so we can use it to optimize control strategy and test new facility and equipment designs. We also work closely with the data science and AI teams at Meta to couple these models with various machine learning methods, such as reinforcement learning, to predict the best set of action points in real time. In addition, we are exploring other new modeling approaches, combining recent advances in physical science and machine learning to find cross-cutting approaches that will further improve the efficiency, reliability and sustainability of our infrastructure.
Meta believes in building community using open source technology. Explore our latest projects in artificial intelligence, data infrastructure, development tools, front-end, languages, platforms, security, virtual reality and more.
Data Center Hvac Cooling Systems
I distinctly remember a conversation with a mechanical engineer who had operated data centers for many years. He felt that most mechanical engineers did not really understand the operation and design of the data center. He explained that most HVAC engineers start out in office or residential design, focusing on comfort cooling, before moving into data center design. He thought the paradigms they were learning in those design projects didn’t necessarily translate well to data centers.
It is important to understand that comfort cooling is not the primary goal of data center cooling systems, although the data center must be safe for the people who work in it. In fact, it is perfectly acceptable (and typical) that areas within a data center are not conducive to long-term occupancy.
Like any good engineer
Pdf) Thermal Metrics For Data Centers: A Critical Review
Data centers energy consumption, energy efficient thermal curtains, energy efficient data, data centers energy, energy efficiency in data centers, energy star data centers, energy efficient data centers, aligned energy data centers, energy efficiency data centers, efficient data centers, energy efficient data center, efficient energy management