Modeling in 3D
December 27, 2005
Appeared in Energy & Power Management
Modern solutions to avoiding data center hot spots
As technological advancement continues to stoke the demand for high-density data centers, the specter of overheating looms large in the background. More than ever, organizations will need comprehensive cooling solutions that evaluate every aspect of the data center.
Without a proper understanding of the cooling and air-flow patterns within a data center, it is not possible to efficiently cool the servers in the space. Typically, companies often add unneeded HVAC systems to their data centers in an effort to prevent the overheating of servers. While this may provide a temporary solution, computational fluid dynamics (CFD) modeling provides a more comprehensive understanding of a data center's heat loads and the cooling enabling the organization to circumvent the need for extraneous and expensive mechanical equipment.
However, traditional CFD modeling often focuses solely on under-floor air distribution, ignoring the air flow above the raised floor. The following sections will evaluate the theory, practice. and execution of 3D modeling in today's high-density environment.
Theory in Practice
Two years ago, two separate financial institutions sought a method to adequately cool equipment that had registered record-high heat densities. This was a direct result of the fast pace of technological advancement within the banking industry. As the influence of technology in the financial world grew exponentially, the need to house more equipment in each data center became a great necessity. While this modern equipment demonstrated superior performance, it ran at higher temperatures, making overheating a core concern.
These two firms approached Syska Hennessy to evaluate how CFD modeling could be used as a tool to ensure that the equipment could be adequately cooled, while maintaining reliability and uptime for these hypercritical facilities. Syska Hennessy concluded that CFD modeling is not a stand-alone solution, but rather, a key part of the equation.
These results led Syska Hennessy to create a solution that would evaluate the many, disparate elements involved in high temperatures to create data centers that are substantially less likely to overheat. Accomplishing this would save organizations thousands of dollars typically expended on excess physical space, downtime, and equipment replacement costs.
In addition to under-floor modeling, Syska Hennessy layered the third dimension of air flow within the room into a CFD model called the Syska High Density Cooling Solution. Syska Hennessy leveraged its extensive knowledge of HVAC systems and overall data center experience to model the data centers in 3-D and determine real-world solutions.
Actual test data measured at real data centers, compared with calculated results from CFD modeling has demonstrated this modeling method to be between 85 and 90% accurate.
Syska Hennessy learned that one organization had installed more than 60% of the needed air conditioning to cool the center. The solution was not connected to overall air condition capacity, but to rack configuration. Re-arranging the rack configuration and evenly distributing the air within the data center drastically reduced the need for more air conditioning.
This particular solution could potentially yield savings of $200,000 up front by eliminating the need to purchase two, 20-ton units, with overall energy savings from running fewer units reaching $7,000 to $10,000 annually.
Many organizations across an array of industries are currently employing one-dimensional cooling solutions that ignore the far-reaching needs of the data center. These band-aid solutions, while logical in practice, often represent an expensive and inefficient approach to a more extensive problem.
For example, many outdated solutions, such as moving floor tiles and adding more HVAC may provide a temporary solution but will not be reliable systems for running today's sophisticated data centers, as hot spots will continue to exist.
Furthermore, using yesterday's CFD modeling alone often amounts to “garbage-in, garbage-out” number crunching, leading to solutions that often fail when implemented.
The true need is for an out-of-the-box solution that maintains the required static pressure under the floor and provides increased flexibility in arranging equipment. This ensures that the entire floor will be pressurized evenly to allow the IT systems to function correctly.
Rather than simply viewing this as a numeric computation, solutions must be employed that combine logistical, math, design, construction and operations factors into the final plan.
This becomes evident when evaluating the change in temperature from rack to rack over years of technological advancement. In the past, when data centers were more sparsely equipped temperature readings varied from 2-5%. Today, racks in real-life data centers may vary in temperature by as much as 30%.
Seeing the Big Picture
The implementation of a 3-D modeling solution enables engineers to step back and view the situation as a whole, employing solutions that address the various heat-causing aspects of each data center.
For example, while many engineers are focusing intently on discharge temperatures, they are ignoring the tremendous heat they are generating at the inlet, which is often the primary culprit. While inlet temperature should not exceed 70 degrees F, we often find inlets in the 80 to 90 degree F range. This is deadly for equipment, where every 10 degree F rise in temperature cuts the life of the machinery in half.
Syska Hennessy also ran its custom CFD models on cabinet types and found that most are installed with air re-circulating in the cabinet, which intensifies temperatures and makes the equipment much too hot. Additionally, many cabinets are inherently inefficient for cooling. By implementing a solution predicated on these findings, these problems can be immediately addressed.
In one data center, the configuration sucking the air back into the rack resulted in inlet temperatures of 90 degrees F, and discharge at 120 degrees F.
In another case, air at 80 cubic feet per minute (cfm) was being delivered to a cabinet instead of 1,000 cfm because their dampers were closed on the tiles. The implementation of new tiles raised the 80 cfm back up to the 600 -800 cfm range, without adding any extra HVAC equipment.
Finally, a data center was losing air flow through the wall openings for the cables. In this instance, the data center was losing 40% of its air through these openings. The temperature remained stable in the room, but the air was not circulated to where it was needed at the front of the cabinets.
These disparate cases, while fundamentally different in practice, all relate to the high importance of using 3D models to evaluate the various potential contributors to high temperatures in data centers. Rather than adding a few more racks to a data center, forward-thinking organizations must use this resource to determine exactly where to place them, with the right amount of cooling, HVAC, and configuration to create an adequate cooling solution. As technology continues to advance, cooling solutions must advance alongside to ensure the maximum reliability and efficiency.
Joseph O'Sullivan has more than six years of experience as a consulting mechanical engineer and has specialized experience focused on HVAC design for critical facilities. A member of Syska Hennessy's Critical Facilities Team, Sullivan is a co-author of the Syska High Density Cooling Solutions and has extensive experience with in-house CFD Modeling of airflows within mission critical facilities.
Jun Yang, PE, is a senior mechanical consulting engineer with Syska Hennessy Group's Critical Facilities team, specializing in the planning, design and construction of Critical 7x24 Facilities. He handles project management and system design for clients with mission critical facilities across the country. In addition, Jun is a key member of the Syska High Density Cooling Solutions Task Force, and currently focuses his efforts working with CFD Modeling and other tools to improve the efficiency of high density critical environments.