In many data centers, traditional room-based cooling systems are reaching the limit of their capabilities. With IT equipment now pushing peak power density to 20 kW per rack or more, the room-based systems – which are designed for power densities on the order of 1-2 kw per rack – just can’t keep up with the heat output.
To address the problem, new design approaches are emerging that focus on integrating air conditioning with specific rows or racks, providing better predictability, efficiency and other benefits as compared to room-oriented systems.
All data center air conditioning systems have two key functions: to provide cooling capacity and to distribute the cool air to the IT loads that need it. The first function is the same no matter what type of cooling system is employed; it’s the second that varies.
Room-oriented cooling architecture
A room-oriented architecture may consist of one or more air conditioners supplying cool air to the entire room. In smaller rooms, racks may be placed with no specific attention paid to airflow. In larger, more sophisticated installations, raised floors may be used to direct air into hot-aisle / cold aisle layouts. Air flow will be affected by the unique constraints of the room, including the ceiling height, room shape, obstructions above and under the floor, rack layout, CRAC location and the distribution of power among the IT loads. As a result, it’s difficult to predict performance, particularly as power density increases and the data center configuration changes. And much of the cooling capacity may be wasted if the cool air never reaches an IT load that needs it.
With a row-oriented architecture, the CRAC units are dedicated to a particular row and are mounted among the IT racks, overhead or under the floor. This results in far shorter, more clearly defined airflow paths as compared to the room systems. In addition, airflows are much more predictable because all of the rated cooling capacity of the CRAC can be directed at IT loads.
The reduction in the airflow path length also reduces the CRAC fan power required, increasing efficiency. It also allows cooling capacity and redundancy to vary by row, with perhaps one row of racks running high density applications such as blade servers, while another row satisfies lower power density applications such as communication enclosures; each gets only the cooling capacity it needs. Same goes for redundancy, with N+1 or N+2 redundancy targeted at specific rows.
The rack-oriented architecture is similar to the row except that the CRAC units are dedicated to and mounted within a specific rack. In this system, airflow paths are even shorter and more exactly defined than either the room or row systems, so much so that they’re immune to any installation variation or room constraints. All of the rated capacity of the CRAC can be utilized, enabling the highest power density (up to 50 kW per rack) to be achieved. The rack-oriented architecture also has the same benefits as the row system in terms of increased efficiency, and highly targeted cooling and redundancy. The main drawback of this approach is that it requires a large number of air conditioning devices and associated piping when compared to the other approaches, particularly at lower power density.
Many data centers will likely use a mix of all three cooling architectures to cover varying requirements. To learn more about each architecture, and where it will best fit into your data center, read the APC by Schneider Electric white paper, “The Advantages of Row and Rack-oriented Cooling Architectures for Data Centers.”
7 years ago
I wonder if anyone else could have put it better! Absolutely second every word written here. Great job.