• At last! The correct way to calculate data center power density

Default Alternative Text
By isolating and documenting space and density requirements, you can ensure a facility will perform predictably.

The historic method of specifying data center power density using a single number of watts per square foot (or watts per square meter) is an unfortunate practice that has caused needless confusion as well as wasting energy and money. But there’s a logical approach that will enable you to improve electrical efficiency and avoid excessive first-time costs.

Understanding the variables

The traditional calculation method yields a statement such as “2,000 square feet at 1,000 watts per square foot.” This presents more questions than it answers. What’s included in the area or power calculation? How does it relate to the number of IT cabinets or devices? What’s the power variation value across IT cabinets? How does the number apply as a data center expands over time?

A better approach to specifying density assumes that IT power varies among cabinets as well as over time, and addresses the issues of modularity and growth.

Utilizing the proper method of specifying density is important. Errors can result in unpredictable performance if you specify too low, or result in high operating expenses if you specify too high.

Try a new approach

What’s the solution? First, consider this:

  • The unit of physical space in the density specification is the IT cabinet rather than floor area.
  • A specification needs to be hierarchical and modular to allow for differing density requirements for rooms and zones.
  • IT cabinets have different power requirements that may not be well-defined in advance and may vary with time.

To illustrate, we’ll focus on a 40 kW server room with a single pod. In this case, the level of specification is the room, which is also the pod, and which contains a group of IT cabinets. You can specify the density parameters for this server room by following this procedure:

  1. 1. Determine the number of cabinets based on the IT requirement. You will convert the per unit values of power, cooling and space to total values.
  2. 2. Determine the design target average power per cabinet based on vendor specifications or by choosing typical average design values for the application to size the bulk power and cooling systems.
  3. 3. Determine peak power by establishing the maximum expected or allowable cabinet power to size power distribution and cooling distribution system requirements.
  1. 4. Estimate power uncertainty by considering different scenarios for IT deployment or by choosing typical design values for the application. This is essential to determine the reserved space needed and avoid stranding costly power and cooling capacity.
  2. 5. Estimate the managed power ratio based on the expected power management functionality of the IT load, establishing the operating load points for power and cooling systems to determine efficiency and energy use.
This procedure applies equally well to server rooms or data centers, whether you’re improving an existing facility or designing a new one. By isolating and documenting space and density requirements, you can ensure a facility will perform predictably.
You’ll find detailed examples and worksheets that will take you through this procedure in White Paper #155, “Calculating Space and Power Density Requirements for Data Centers.”
 
SPECIFICATION PARAMETER WITH DEFINITION AND HOW IT IS USED IN THE DESIGN

1. Number of units (#)
  • Number of cabinets in a pod, pods in a room, or rooms in a facility.
  • To convert the per unit (per cabinet, per pod, per room) values of power, cooling, and space to total values
2. Design target average power per unit (kW)
  • Expected full load (rated) power per unit averaged across the population
  • To size the bulk power and cooling systems for the level (pod, room, facility)
3. Peak power per unit (kW)
  • Maximum expected power of the highest unit in the population
  • To size power distribution and cooling distribution system requirements
4. Unit power uncertainty (%)
  • Quantifies the expected uncertainty of the actual power compared to the design target average power
  • To determine the reserved space needed to ensure that low density deployment will not strand costly power and cooling capacity
5. Managed power ratio (%)
  • Power reduction factor (% of design target average) due to power management functions in IT equipment
  • To establish the operating load points for power and cooling systems to determine efficiency and energy use

For examples and worksheets, download the complete White Paper #155.
Learn more