Data centres become carbon tax scare-story
Your rackspace will cost you more!
Given that the IT business swallows nearly three percent of Australia’s power output, it’s hardly surprising that the industry is viewing the country’s coming carbon tax with some trepidation .
All IT users should at very least count up their kilowatt-hours and look at actually using the power-saving technologies that come free with most hardware, since Connection Research says the aggregate power demand of Australia’s computers is as much as seven percent of total electricity consumption.
With almost dead-boring predictability, the advent of the carbon tax has even led to the nearly-xenophobic prediction  that our data will move offshore en masse (Australians keep pretending we’re internationalized and cosmopolitan, an image we drop like a live grenade the instant someone says “business will move offshore”).
It’s true that electricity is the main input to the data centre business, and at more than 18 percent of ICT emissions, (Connection Research) of the country’s power, it’s a power hog. Its consumption is outsized compared to the economic scale of the industry.
Back in February, Pacnet predicted that the combined data centre markets of Australia and New Zealand would reach $AU1.1 billion by 2016. When it reaches that point, it will be around 0.1 percent of Australia’s trillion-dollar GDP: we could shift the entire data centre business offshore next week and barely notice.
But that’s not going to happen: some data will remain here because it will be illegal to shift it offshore, and some data centres will adjust, because good design and intelligent management can slash the data centre power bill.
So why aren’t we already there? Again calling on data from Connection Research, we devote 2.7 percent of total power generation to ICT, compared to an international norm of around 2 percent, and a fair slice of that relative inefficiency is probably attributable to data centres.
One reason is a simple lack of incentive: until recently, electricity was relatively cheap, and that influenced data centre design. It was cheaper to use more power than to buy better technology.
Power prices were already rising
It’s worth noting that electricity prices have already been rising sharply, particularly in NSW. Those rises (as much as 18.1 percent in 2011 alone), attributed to transmission infrastructure upgrades rather than environmental policy, far outrun the ten percent price rise predicted when the carbon tax comes into effect.
On that score, the idea that the carbon tax is going to force data offshore is already debunked: the price rise for commercial customers (who have negotiating power) is unlikely to match the household ten percent price rise, and much larger rises in recent years haven’t sent our data to Singapore. There has, in fact, been a continuing data centre boom in the face of those price increases.
Something that hides underneath the shiny public image of the data centre industry is that most data centres are fairly old (and therefore inefficient), and a great many of them aren’t purpose-built wonders of civil engineering, but are small rack-rooms in cheap real estate.
That probably hints at where the hammer will fall: if a small data centre combines inefficiency today with a lack of capital for upgrades, it will suffer more than a newer facility that’s backed by sufficient capital to pursue energy efficiency.
Age shall weary them
Another consideration is the age of Australia’s data centre stock. Here again, perceptions are skewed by the media attention given to new data centres, and we forget just how many of today’s data centres are retrofitted mainframe rooms from the 1960s and 1970s.
Until the last few years, there wasn’t much reason to retrofit such facilities for energy efficiency: the presence of big power feeds, proximity to carrier fibre, and existing air-conditioning ducts made a given site attractive, because a retiring mainframe room made a cheap, quick – and dirty – site to run up a new data centre.
I’m not aware of any inventory of data centre stock by age, but I would not be surprised to hear that half of today’s data centres are in facilities more than 20 years old, and have inherited inefficient power and cooling designs.
Those data centres also have a lot to lose: retrofitting an existing building is much more expensive than designing efficiency into a new one, especially if nobody really gave thought to cooling or power consumption in the original design.
Even that’s not necessarily a bad thing. Another common characteristic of many of the country’s older data centres is a CBD location (not all; I’m well aware of the exceptions). In CBDs, data centres are already a problem: the next person to try and host a new data centre in the centre of Sydney will find it much harder to get multiple power feeds than was the case ten years ago.
CBD locations also suffer from high real estate costs compared to outer suburban industrial parks, and various other engineering issues.
Historically, telecommunications networks and staffing demands were the twin attractors of data centres to inner-city locations: that’s where you could get the fibre connection; and a CBD was the hub of public transport for all those staff you needed to run a mainframe room.
Both of these have long lost their relevance: there’s much more suburban fibre than there was 15 years ago, and data centres have far lower on-site staff needs than they used to.
On the whole, it seems more likely to me that a new, energy-efficient and well-connected data centre somewhere in Australia will be a far more attractive proposition than a move offshore. It’s quite likely that the cost of a data audit to try and distinguish what can and cannot be moved offshore will be far more daunting for a large enterprise than whatever incremental cost the carbon tax brings to the data centre. ®