Use Smart Cooling to Keep Your Cloud Cool: 4 Innovative Approaches to Hyperscale Cooling
For hyperscale data centers, cooling is a perennial strategic challenge. Keeping the facilities at the right temperature takes a lot of energy, which drives up costs and increases greenhouse gas emissions.
Sponsored by Dell Technologies
Cloud computing is hot. That's not just figurative. Not only is cloud computing one of the hottest technology trends, it also quite literally generates a lot of heat.
An average-sized data center generates between 20 and 50 MW of heat energy, while some create up to 100 MW -- enough to power 80,000 homes. In recent years, chip makers have made great strides toward engineering their CPUs to be more efficient, competitive, and compliant, meaning that each server generates less heat than in the past. But at the same time, greater use of artificial intelligence (AI), machine learning (ML), and advanced analytics is increasing demand for compute power. As a result, data centers are deploying more servers with more processing power, which offsets the efficiency gains.
For data centers, all that heat is a problem. According to the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE), the ideal temperature for a hyperscale environment is between 73 and 75 degrees Fahrenheit, and if they get above 82 degrees, it could damage the servers.
In order to stay within those thresholds, data centers divert a lot of power into their cooling systems. In fact, Gartner determined that cooling and power systems account for 39% of the power consumption in a typical data center. That's just a hair less than the servers themselves use -- 40%.
Using so much power for cooling has a big impact on a data center's bottom line -- and an even bigger impact on the environment. Data centers now account for 3% of global energy consumption. An MIT paper reported, "The Cloud now has a greater carbon footprint than the airline industry." It added, "The electricity utilized by data centers accounts for 0.3 percent of overall carbon emissions."
Those carbon emissions are having a compounding effect on data centers. As the earth's average temperature warms, data centers are experiencing more high-temperature days, which in turn, makes it more difficult to keep their facilities cool.
To counter this effect, governing agencies around the world are considering and/or putting into place legislation aimed at curbing data center greenhouse gas emissions. And others are offering incentives for organizations that reduce their carbon footprint.
The combination of high costs, regulation, and a desire to reduce their environmental impact is driving cloud service providers and other hyperscale data center operators to look for innovative ways to optimize their cooling. Here are four that are worth a closer look:
1. Hot/Cold Aisle Configurations
Hot and cold aisles in the data center are nothing new. Organizations have long relied on these layouts to help cool the data center more efficiently. This configuration enhances the ability of the servers to draw in air from the cold aisle and push it out the back. The direct path reduces hot air and cold air mixing, making cooling systems more efficient. It also improves performance and can extend the lifespan of the hardware.
Some of the latest server hardware included in Dell's Power Edge Scale servers, offer better serviceability of this footprint-conserving layout. The I/O is located at the front of the server -- in the cold aisle. It makes management and cabling much easier and safer for the people maintaining the equipment.
2. Liquid and Immersion Cooling
Liquid cooling has been around since the 1960s, but it was somewhat unpopular because of concerns about costs, complexity, and potential damage from leaks. However, interest in this approach has surged in recent years. New types of liquid cooling can reduce power consumption up to 50%, which can more than offset the costs of installing a liquid cooling solution. The new systems also minimize the likelihood of leaks and quickly alert staff to any problems, which have always been worries about this approach. For example, PowerEdge systems management offers “leak sense” as a part of its capabilities to quickly alert IT administrators automatically if a leak is detected.
In addition to traditional liquid cooling, today's hyperscale operators also have the option of immersion cooling. In these deployments, the entire system is submerged in liquid, which captures 100% of the heat and eliminates the need for air conditioning. That's a huge savings, both in terms of dollars and greenhouse gases, making this approach more attractive for cloud services providers.
3. Advanced Air Cooling
Organizations that aren't ready to take the plunge with liquid cooling also have improved options for air cooling. For example, Dell's Multi-Vector Cooling uses an array of thermal and power sensors to monitor the state of the data center. Algorithms control zoned fans to properly balance the temperature and maximize efficiency.
4. AI Simulation and Systems Management Tools
In an interesting twist, the AI that is driving up power consumption in the data center is also helping hyperscale environments cool themselves more efficiently. Management software with advanced algorithms and predictive analytics capabilities can help operators allocate power more effectively to reduce stranded power. In one case, a cloud provider reduced its costs by 40% by relying on AI models.
Flexible, open management tools are also part of the solution. They make it possible to monitor power usage, carbon emissions, and thermal health. They can implement dynamic cooling strategies and mitigate risks from power and cooling events. They also allow you to set policy-based controls.
New software options combined with infrastructure planning tools give operators the ability to strategize from deployment through lifecycle management.
Today, hyperscale operators have more options than ever to help them keep their cool. And that can help them reduce operating costs along with their carbon footprint. To learn more about new technologies available to hyperscale providers, see the white paper titled, "Don't let the challenges of scale-up keep you down."
About the Author
You May Also Like
2024 InformationWeek US IT Salary Report
May 29, 20242022 State of ITOps and SecOps
Jun 21, 2022