As the business requirements of enterprise companies change, so too do the demands and challenges placed on their IT infrastructures. Today’s data center manager is challenged by a wide array of data center objectives, most notably capacity issues. The 2013 Data Center Users’ Group survey revealed that on average, respondents reported using only 55% of their cooling capacity, suggesting most organizations may be able to get more from their existing infrastructures. Better capacity planning is key.
How IT departments approach their physical infrastructure strategies can affect their effectiveness in balancing these objectives as technologies and business needs transform. To address these demands, IT departments have refocused their approaches to data center thermal management to maximize efficiencies in infrastructure design/deployment, operations, management and planning. For these new strategies to be successful from a capacity and efficiency standpoint, a couple fundamental best practices must be observed.
Best Practice 1: Maximize the return temperature at the cooling units to improve capacity and efficiency
Maintaining appropriate conditions in the data center requires effectively managing the air conditioning loop comprising supply and return air. The laws of thermodynamics create opportunities for computer room air conditioning systems to operate more efficiently by raising the temperature of the return air entering the cooling coils.
This best practice is based on the hot-aisle/ cold-aisle rack arrangement (Figure 1), which improves cooling unit performance by reducing mixing of hot and cold air, thus enabling higher return air temperatures. The relationship between return air temperature and sensible cooling capacity is illustrated in Figure 4. It shows that a 10°F increase in return air temperature typically results in a 30% to 38% increase in cooling unit capacity.
The racks themselves provide something of a barrier between the two aisles when blanking panels are used systematically to close openings. However, even with blanking panels, hot air can leak over the top and around the sides of the row and mix with the air in the cold aisle. This becomes more of an issue as rack density increases.
To mitigate the possibility of air mixing as it returns to the cooling unit, perimeter cooling units can be placed at the end of the hot aisle as shown in Figure 1. If the cooling units cannot be positioned at the end of the hot aisle, a drop ceiling can be used as a plenum to prevent hot air from mixing with cold air as it returns to the cooling unit. Cooling units can also be placed in a gallery or mechanical room.
In addition to ducting and plenums, air mixing can also be prevented by applying containment and by moving cooling closer to the source of heat.
Optimizing the Aisle with Containment and Row-Based Cooling
Containment involves capping the ends of the aisle, the top of the aisle, or both to isolate the air in the aisle (Figure 3). Cold aisle containment is favored over hot aisle containment because it is simpler to deploy and reduces risk during the event of a breach of the containment system. With hot aisle containment, open doors or missing blanking panels allow hot air to enter the cold aisle, jeopardizing the performance of IT equipment. In a similar scenario with the cold aisle contained, cold air leaking into the hot aisle decreases the temperature of the return air, slightly compromising efficiency, but not threatening IT reliability.
Row-based cooling units can operate within the contained environment to supplement or replace perimeter cooling. This brings temperature and humidity control closer to the source of heat, allowing more precise control and reducing the energy required to move air across the room. By placing the return air intakes of the precision cooling units directly in the hot aisle, air is captured at its highest temperature and cooling efficiency is maximized. The possible downside of this approach is that more floor space is consumed in the aisle. Row-based cooling can be used in conjunction with traditional perimeter-based cooling in higher density “zones” throughout the data center.
Supplemental Capacity through Sensible Cooling
For optimum efficiency and flexibility, a cooling system architecture that supports delivery of refrigerant cooling to the rack can work in either a contained or uncontained environment. This approach allows cooling modules to be positioned at the top, on the side or at the rear of the rack, providing focused cooling precisely where it is needed while keeping return air temperatures high to optimize efficiency.
The cooling modules remove air directly from the hot aisle, minimizing both the distance the air must travel and its chances to mix with cold air (Figure 4). Rear-door cooling modules can also be employed to neutralize the heat before it enters the aisle. They achieve even greater efficiency by using the server fans for air movement, eliminating the need for fans on the cooling unit. Rear door heat exchanger solutions are not dependent on the hot-aisle/cold-aisle rack arrangement.
Properly designed supplemental cooling has been shown to reduce cooling energy costs by 35% to 50% compared to perimeter cooling only. In addition, the same refrigerant distribution system used by these solutions can be adapted to support cooling modules mounted directly on the servers, eliminating both cooling unit fans and server fans.
Best Practice 2: Match cooling capacity and airflow with IT loads
The most efficient cooling system is one that matches needs to requirements. This has proven to be a challenge in the data center because cooling units are sized for peak demand, which rarely occurs in most applications. This challenge is addressed through the use of intelligent cooling controls capable of understanding, predicting and adjusting cooling capacity and airflow based on conditions within the data center. In some cases, these controls work with other technologies to adapt cooling unit performance based on current conditions (Figure 5).
Intelligent controls enable a shift from cooling control based on return air temperature, to control based on conditions at the servers, which is essential to optimizing efficiency.
This often allows temperatures in the cold aisle to be raised closer to the safe operating threshold now recommended by ASHRAE (max 80.5°F). According to an Emerson Network Power study, a 10° increase in cold aisle temperature can generate a 20% reduction in cooling system energy usage.
The control system also contributes to efficiency by allowing multiple cooling units to work together as a single system utilizing teamwork. The control system can shift workload to units operating at peak efficiency while preventing units in different locations from working at cross-purposes. Without this type of system, a unit in one area of the data center may add humidity to the room at the same time another unit is extracting humidity from the room. The control system provides visibility into conditions across the room and the intelligence to determine whether humidification, dehumidification or no action is required to maintain conditions in the room at target levels and match airflow to the load.
For supplemental cooling modules that focus cooling on one or two racks, the control system performs a similar function by shedding fans based on the supply and return air temperatures, further improving the efficiency of supplemental cooling modules.
Capacity management is an essential part of the planning and operation of efficient data centers. In the cooling system, traditional technologies now work with newer technologies to support higher efficiencies and capacities. Raising the return air temperature improves capacity and efficiency while intelligent controls and high-efficiency components allow airflow and cooling capacity to be matched to dynamic IT loads.