Computational Fluid Dynamics (CFD) Optimization

Energy-Efficient Data Centers

CyrusOne is committed to institutional integrity and ethics throughout our organization.  We ensure the highest standards through a number of proven controls.

Photo of a tree in a meadow with the sun shining behind it.

Computational Fluid Dynamics (CFD) Optimization

We use Computational Fluid Dynamics (CFD) modeling to simulate the flow of chilled air throughout a facility as one way we design more sustainable data centers. CFD Models are advanced mathematical simulations that require expert configuration and hours on high-performance computers to complete but give key insights into how air and heat move through our facilities.

Chilled air enters the data halls through ventilated tiles into cold aisles created by barriers that surround the servers. The ventilated tiles have different sizes of openings which allow us to control the amount of air that enters each cold aisle, and the barriers keep the chilled air contained to the space where cooling is needed (near the air intakes for the servers). These barriers can include blanks installed in empty racks, end-of-row doors, and roof panels (rack top baffles). Together these features help optimize the amount and location of cooling while reducing the overall power demand of the system. When this arrangement is not working properly, chilled air is directed into equipment that does not require as much cooling while missing other areas that need cooler air. The air handler fans then need to work harder to meet the supply air setpoint, wasting electricity. We adjust our cooling through CFD modeling to best support our customers’ evolving needs.

Optimizing the speed of air handler fans can yield surprising results. Fans consume more electricity to turn faster, but electricity consumption doesn’t increase linearly. For example, we might expect a fan running at 100% speed to use four times as much power as one that is running at 25% speed. In fact, it uses 30 times more power since air resistance against the fan blade increases geometrically with increased speed. So, three fans moving at 33% speed will move the same air as one fan running at 100% speed, but they will use 86% less power. This is key to understanding how CFD modeling can achieve significant energy savings by fine-tuning our cooling performance.

This is one way we address the unique challenges of a colocation data center environment. Unlike in-house data centers, colocation data centers have a split responsibility between the servers (controlled by the customers) and the cooling systems (controlled by CyrusOne). Coordinating these two efforts for energy efficiency is not a simple matter. By using CFD modeling we can recommend optimal settings for our cooling equipment and customer server arrangements to ensure that both operate efficiently. CFD Modeling provides opportunities to customize the cooling of each data hall between construction and operation, during customer build-out, and for ongoing optimization.

Between Construction & Operation

During commissioning, data centers are tested to ensure that the cooling system can remove the heat that will be generated by servers. The data centers must remain running after being tested so that they are fully operational when customers move in. Unlike in-house data centers, colocation data centers face the uncertainty of not knowing exactly when customers will install their equipment, so there is often a period when the data center support systems are running without any servers occupying the space. Using CFD modeling, we have optimized cooling during that fallow period by reducing fan speeds and strategically allocating ventilated floor tiles, thereby lowering total energy demand.

Supporting Customer Build-out Process

Our colocation data centers are designed to flexibly use cooling system configuration and cold aisle containment to match the varying need for cooling based on different data hall capacities and occupancy rates.

Modeling Cold Aisle Containment
Poor cold aisle containment lets precious chilled air spread around the data hall, rather than efficiently cooling the servers Good cold aisle containment delivers chilled air to the servers where it is needed, saving energy
Images created by CyrusOne using TileFlow by Innovative Research Inc.

When customers move in and begin installing servers in our data halls, we bring the expertise of our CFD modeling to make recommendations on server arrangement, cold aisle containment, and chilled air flow rates to maximize the efficiency of both our equipment and theirs. This is the beginning of our partnership for efficiency with our customers.

Ongoing Optimization

After customers have moved in, we continue to look for opportunities to improve efficiency in our effort to have more sustainable data centers. Data halls evolve over time: servers get upgraded, their loads change, and their temperature tolerances change. If we don’t evolve our cooling strategies with them, then our cooling systems gradually become less efficient. We run the CFD modeling periodically to inform updates to our cooling system arrangement (such as optimizing airflow tiles and cold aisle containment) so that we can maintain efficiency without affecting customer server placement.

Source: Product images from Upsite Technologies, Inc. and Polargy, Inc.

Source: Product image by STULZ Air Technology Systems, Inc.

 

 

Source: product images by Tate Access Floors, Inc.