Comments on: Data Centers – Where Did All the Water Go?

By: Andy Lawrence

Andy Lawrence — Tue, 07 Jul 2009 11:16:25 +0000

This is an interesting suggestion, but some way ahead of its time. In many geographies and organizations, water use isn’t even metered – that would be a start. Once we have a metric, we would then have a complicated debate – a good metric in power leads to a poorer metric in water, or vice versa. Managers would probably end up with a formula based on financial savings, rather than what is best for the planet or even for the local neighborhood. That isn’t what the EPA or the US intended. Good debate ahead, though.

By: Ian Bitterlin

Ian Bitterlin — Tue, 07 Jul 2009 09:22:20 +0000

Use of water in data-centres is a very North American centric bad-habit. In EMEA (largely for reasons of the health hazards arising from water towers) we tend to use more air-cooled chiller plant in data-centres. The link between data-centres and health is that a local case of legionella can cause a water tower to be immediately shut down by many local health authorities – not condusive to uptime!

By: Jud Cooley

Jud Cooley — Tue, 07 Jul 2009 02:53:13 +0000

As it was with PUE, the most immediate value of a metric of any kind is that it gets people to start measuring consumption. If water use is measured it can be managed. I think it is premature, however, to think of this in terms of a productivity metric. The industry is currently struggling with which productivity metric (or proxy) is the best one to use and, until that argument settles out, metrics will be used inconsistently and to game the system. Since the water in question is used to supply cooling, a usable metric will relate water use to either the amount of cooling provided or the amount of electrical power use that is subject to that cooling. Comparing quantity of water used to either the numerator or denominator in the PUE metric would make sense. Gallons of water per watt of power delivered to it’s point of use.

By: ThinkThank

ThinkThank — Mon, 06 Jul 2009 20:12:35 +0000

Your suggestion of measuring water consumption soley against compute power is misguided. Unless it takes in account PUE or power consumption of the mechanical systems it will only lead to inefficiency. Direct expansion cooling systems consume several times more power per ton of cooling than evaporative units. I agree we need to be measuring and tracking water consumption of large (>1mw) datacenters. I believe your suggestion of measuring against compute cycles is self serving as Liebert is not competitive with McQuay, Trane, and Carrier for large deployments.