Why large data centres are not always more efficient

Everyone say big data centres are more efficient. It’s true, until it isn’t. Then things can get much worse.

Last week, I wrote about how larger data centres offer better efficiency than smaller ones due to economies of scale.

But massive data centres are not always more resource efficient. I highlight two real-life cases.

When decisions get forced

A industry contact recently told me about a Johor data centre that resorted to air-cooled chillers instead of water-based evaporative systems.

While this saves substantial water due to reduced water loss, such chillers consume more energy, resulting in a higher PUE an engineer estimated at up to 1.5.

The theory was that the facility couldn’t secure adequate water supply, unlike another data centre in the vicinity.

Unfortunately for a hyperscale data centre, this means the unfavourable PUE gets locked in for its entire operational lifespan.

PS: Chillers are typically located on the roof of data centres in Johor, so there’s no hiding them.

When inexperience meets hyperscale

Fresh from my visit to Schneider Electric’s factory in Barcelona, I related the possibility of microclimates forming around data centres to a technical data centre leader.

To my surprise, he shared an anecdote of touring an almost-completed data centre where he spotted a massive design error involving this issue.

• Chillers placed too close together.
• Resulting ambient temperature too high.
• When operational, they won’t operate optimally.

He didn't know if the team rectified the mistake, though he conceded it was unlikely given the advanced state of construction.

Bigger doesn't always mean better

So there you go. Larger data centres can be more resource efficient, but this assumes proper design and implementation.

On the other hand, whatever mistakes made by a hyperscale data centre operator gets amplified on a far larger scale.

Do you have horror stories of data centre built wrong? Leave a comment or drop me a direct message.