The Top 9 Mistakes in Data Center Planning

Overview

The authors argue for a total-cost-of-ownership model combining capital expense, operations and maintenance expense, and energy cost, with a 3-7 year O&M budget needed to build an ROI model. Design criteria should come from business goals and risk profile before a tier, site or space plan is chosen. The paper stresses that not everyone who wants a Tier 3 design needs one, that a 'must have 300 watts per square foot' requirement is often not justified by business needs, and that higher tiers raise O&M and energy cost.

On obsolescence, it cites IDC for an average data centre age of 9 years and Gartner for treating any site over 7 years old as obsolete. On space, in the most robust systems the ratio of raised floor to support (mechanical/electrical) space can be as high as 1 to 1, and office, equipment yards and staging areas are often overlooked. On flexibility, designs should allow adding UPS capacity to existing modules without an outage and uninterrupted addition of in-rack or in-row cooling for future high-density rollouts.

PUE claims must state the design day (perfect day versus yearly average) and the load level. Water is a separate line item: a typical 2 megawatt data centre using water-cooled towers needs 50 to 60,000 gallons of make-up water per day. Other pitfalls are underestimated build budgets, inflexible dead-end designs, treating LEED as an afterthought, and over-complex redundancy schemes that add failure points and human-error risk.

The paper is qualitative but contains a few useful planning ratios: PUE context, the support-space ratio and the make-up water figure. The numbers rest on one 2 MW example and cited analyst views, so they should not be read as universal thresholds.

Key points

1. IDC puts the average data centre at 9 years old; Gartner calls any site over 7 years obsolete.
2. TCO combines capital expense, O&M expense and energy cost; a 3-7 year O&M budget is needed for an ROI model.
3. Not everyone wanting Tier 3 needs it; '300 watts per square foot' is often unjustified; higher tiers raise O&M and energy cost.
4. In the most robust systems raised floor to support space can be as high as 1 to 1; offices, yards and staging areas are often overlooked.
5. Designs should allow adding UPS capacity to existing modules without outage and uninterrupted addition of in-rack/in-row cooling.
6. PUE claims must state the design day (perfect day vs yearly average) and load level.
7. A typical 2 MW data centre with water-cooled towers needs 50-60,000 gallons of make-up water per day.
8. Other mistakes: underestimated build budget, dead-end design, LEED as afterthought, over-complex redundancy.

Using it in design

• Require design criteria (tier, density, capacity) to be derived from stated business goals before site and space tools are used.
• In the space planner, estimate electrical and mechanical support space up to 1:1 with white space for high-redundancy designs, plus office, yard and staging areas.
• Always label PUE with its load fraction and whether it is annualized or a design-day value.
• For cooling-tower designs show daily make-up water beside PUE (roughly 25-30k gal/MW/day from the 2 MW example); build the TCO view as CapEx + multi-year O&M + energy, and warn when redundancy schemes add components without availability gain.

Who should read it

Owners, investors and design managers planning new or expanded facilities in the 1-3 MW range.

Related standards

• ISO/IEC 30134-2
• LEED

Publication details

Authors

Mike M. Hagan; John Lusky; Tuan Hoang; Scott Walsh

Publisher

Schneider Electric

Year

2012

Document number

WP 145

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation