G00218031

A Simplified Data Center Sizing Model

Published: 14 November 2011

Analyst(s): David J. Cappuccio

Before deciding how large a data center should be, a logical model should
be developed to show how much growth any given space should
accommodate. That growth should include not only the compound annual
growth rate (CAGR), but also the impact of technology refresh cycles and
increased compute density. You may be startled by how much less space is
needed to accommodate expected growth.

Key Findings

Defer capital expenditures first by becoming more efficient in the current space.

At 80%-plus utilization, develop plans to consider alternatives or coping strategies.

Increase the technical refresh cycle as a way to increase capacity and defer data center
upgrades.

Recommendations

Servers are becoming less hungry for power as manufacturers focus on power reduction.

Optimize rack density more fully to extend the life of data centers.

Increasing the density of space often moves the problem from space to energy as the
overriding bottleneck.

Analysis
When looking at data center design, or even retrofitting an existing data center, the question of size
always comes up first (or close to it). IT staff have always looked at data center sizing as something
they started with too much of, and slowly grew into until they had too little, at which point the sizing/
planning exercise began all over again.

Data Center Capacity Planning

Traditional long-range capacity planning was done at the data center level essentially estimating
future floor space requirements based on historical growth rates, and rarely taking into account how
efficiently the current floor space was utilized. This method has been used for years to size out the
next-generation data center, and, in many cases, resulted in companies building overprovisioned,
oversized and overpriced data centers.
Getting into detailed capacity planning for data centers can be a daunting task, especially when all
factors are taken into account. At a macro level, things like business climate, historical growth
trends, and merger and acquisition (M&A) activity projections need to be factored in and from a
strictly IT perspective, it gets even more speculative. Server and storage growth estimates need to
map to technology release schedules, refresh cycles, space requirements and performance
estimates. And for a clear picture, all of this needs to map to estimated power consumption and
cooling requirements over time, to get a high-level view of what type of facility might be needed,
and how much space should be planned for.

Plan for Compute Requirements, Not Capacity

For a faster approach, consider breaking down the IT estimates based on the servers, or compute
units (CUs), in use today, and capacity estimates based on expected market trends. This can create
a baseline for more-detailed thinking and analysis. What is a CU? It could be anything you decide;
however, for this example, let's base it on a physical device in a rack. For example, if a standard
server today averages 2 units (U) in size, each 2U within a rack equates to 1 CU. Rack-based
storage and networking equipment could be sized using the same method, just with different U
allocations. The reason we use U as a standard mechanism is for consistency in space planning (XU
needed today plus 10% growth = YU tomorrow), and because future technologies will all use a
variation on the U for sizing. A single CU may take up 2U, but next generations of servers will
require only 1U, or even .50U or .25U, thus increasing the potential capacity of a rack.
Let's assume an environment today with 40 racks with an average device density per rack of 60%.
These racks are designated for servers, and are nearing capacity limits of 504 physical devices.
From a capacity and space-planning perspective, there are multiple alternatives to consider: scaling
upward within the racks with current technology; scaling via new technology; scaling outward by
acquiring more space for racks; or a combination of scale-up and new technology.
The current environment of 504 servers (or CUs) is allocating 1,008U within the racks (40 racks x
42U x 60% capacity). By upgrading to newer 1U servers, the CU count is now doubled to 1,008
servers, while the rack space remains the same. Power density has increased, but not dramatically
since next generation servers are less power hungry, and performance has increased by at least a
factor of four, due to increases in the number of cores, and faster processor and memory speeds.
While upgrading a single generation of systems in place is a standard practice for increasing both
performance and efficiency, when applying the same logic to upgrades across multiple generations
and form factors, the potential impact on space planning can be dramatic.
For example, applying this logic to the next few generations of servers results in the chart depicted
in Figure 1. The green bar on the far left reflects today's capacity limit on the 40-rack environment.

Page 2 of 7

Gartner, Inc. | G00218031

The total compute units per square foot is 0.42 (assuming 30 square feet per rack), and the total
number of servers (or CUs) is 504. With upgrades in place to 1U, and then to side-by-side servers (.
50U each), and eventually to .25U servers (e.g., skinless), the CU-per-square-foot ratio jumps to
6.72 and performance increases by more than 700%, while the total physical server count reaches
over 4,000, all within the same 40-rack environment.
Figure 1. Growth Potential Within a 40-Rack Environment
CUs per Sq. Ft.
8

Units
9,000
8,000

7,000

6,000

5,000
4
4,000
3

3,000

2,000

1,000
0

0
2U Servers

1U Servers
Performance/Capacity

Side by Side
CUs

Quad Tray

CUs per Sq. Ft.

Source: Gartner (November 2011)

While this growth sounds impressive, there are often negative cascade effects to increased density
namely, power, cooling and cabling issues arise. Figure 2 is an estimate of the power
requirements that a rack would need with this upgrade strategy. Through the second generation of
upgrades, the power density remains manageable for most data centers (upward of 6 kilowatts [kW]
per rack); however, when .50U devices are installed, the power density jumps to 17 kW per rack,
which may be problematic for an existing data center, but becomes a design consideration for new
ones.

Gartner, Inc. | G00218031

Page 3 of 7

Figure 2. Kilowatt Growth Per Rack

Density
8

kW
25

7
20

6
5

15

4
10

3
2

1
0

0
2U Servers

1U Servers
kW/Rack

Side by Side

Quad Tray

Compute Density Sq. Ft.

Source: Gartner (November 2011)

The above two examples are illustrative of what kinds of capacity can be seen using existing racks,
and not increasing the average rack density beyond 60%. This scenario allows for significant
growth within the same footprint, but also mitigates some of the power and cooling risk by
minimizing the density of individual racks.

Traditional Techniques Often Overprovision

Using the more traditional finite approach to capacity planning would yield significantly different
results (and space requirements) than what we have discussed above. For example, assuming the
same 40-rack environment with 60% rack density on average, with 15% growth on average each
year, during a seven-year time frame, this would project out a total server density of almost 1,400,
requiring approximately 3,400 square feet of floor space (see Figure 3).

Page 4 of 7

Gartner, Inc. | G00218031

Figure 3. Traditional 15% Growth Estimates

Square Feet
3,500

Density
1,600
1,400

3,000

1,200

2,500

1,000
2,000
800
1,500
600
1,000

400

500

200
0

0
2011

2012

2013

2014
Servers

2015

2016

2017

2018

Floor space

Source: Gartner (November 2011)

This horizontal growth strategy, while well-understood and historically based, often requires far too
much space, and subsequent capital outlay to justify continuing its use. Comparing this growth with
the options outlined in Figures 1 and 2, it is clear that a 15% growth strategy can easily be
accommodated within the original confines of the 40-rack environment, and well beyond the sevenyear horizon. If managed properly, this strategy provides significant growth within the existing data
center, but also delays the requirement for new data center space for many years to come. This
capital deferment strategy can be used as a strong argument when developing a business case for
faster technology refresh cycles within IT.

Bottom Line
When in the early stages of data center capacity planning, first look to capital deferment, not floor
space growth, as a planning tenet. Focus on ways to optimize what you have first, and then design
for continuous optimization via technology refresh cycles (vertical growth), instead of using
traditional horizontal growth techniques.

Recommended Reading
Some documents may not be available as part of your current Gartner subscription.
"Now Is the Time: Replace Servers Early and Save Money"

Gartner, Inc. | G00218031

Page 5 of 7

"$29 Million Capital Deferral Program: Two Data Center Efficiency Examples"

Page 6 of 7

Gartner, Inc. | G00218031

GARTNER HEADQUARTERS
Corporate Headquarters
56 Top Gallant Road
Stamford, CT 06902-7700
USA
+1 203 964 0096
Regional Headquarters
AUSTRALIA
BRAZIL
JAPAN
UNITED KINGDOM

For a complete list of worldwide locations,

visit http://www.gartner.com/technology/about.jsp

2011 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered trademark of Gartner, Inc. or its affiliates. This
publication may not be reproduced or distributed in any form without Gartners prior written permission. If you are authorized to access
this publication, your use of it is subject to the Usage Guidelines for Gartner Services posted on gartner.com. The information contained
in this publication has been obtained from sources believed to be reliable. Gartner disclaims all warranties as to the accuracy,
completeness or adequacy of such information and shall have no liability for errors, omissions or inadequacies in such information. This
publication consists of the opinions of Gartners research organization and should not be construed as statements of fact. The opinions
expressed herein are subject to change without notice. Although Gartner research may include a discussion of related legal issues,
Gartner does not provide legal advice or services and its research should not be construed or used as such. Gartner is a public company,
and its shareholders may include firms and funds that have financial interests in entities covered in Gartner research. Gartners Board of
Directors may include senior managers of these firms or funds. Gartner research is produced independently by its research organization
without input or influence from these firms, funds or their managers. For further information on the independence and integrity of Gartner
research, see Guiding Principles on Independence and Objectivity.

Gartner, Inc. | G00218031

Page 7 of 7