If Liam Newcombe, secretary of the British Computer Society data centre group, intended to ruffle a few feathers at this year’s Future of the Data Centre Conference, then he was, by any measure, extremely successful.
Newcombe began with a frank appraisal of the current state of the UK’s data centre landscape; it was not complimentary. The combination of business demands and an onerous regulatory environment was heaping pressure on the already overburdened corporate data centre. “We have a serious problem,” he told the audience, but so far the responses to that problem had simply exacerbated the situation.
The perpetration of “incredible inefficiencies” has brought the data centre industry to the brink of crisis, says Newcombe. Those inefficiencies have largely stemmed from the tendency to over-specify data centres in a bid to ensure both high availability and increased resilience – much of which, it turns out, is unnecessary. Most data centre managers, argues Newcombe, “are not aware of how reliable their data centres need to be; they are just frightened that it might not be reliable enough.”
These fears have prompted many organisations to build out hefty, energy-hungry, tier-four facilities, when in reality less heavy-duty infrastructure would have sufficed. Worse still, the most resilient data centres are actually the most inefficient, ensuring that energy consumption is far higher than need be. Energy use within the data centre had doubled between 2000 and 2005. “There is no reason to suspect that this trend is not going to continue. It may even get worse,” says Newcombe.
The IT industry has been aware of the financial and environmental impact of this accelerated energy consumption for some years, Newcombe says, but has largely failed to address the problem. “Frankly, I’m a little surprised that we don’t have environmental demonstrators protesting outside data centres.”
Nevertheless, Newcombe foresees ways round the problem – providing the IT industry is prepared to employ a little imagination. “Forecasting what the data centre of five to 10 years will need to support it is impossible”, but there are several provisions – immediately deployable – that would serve to partially alleviate the problems of space, energy, and cooling.
Where energy demands are concerned, onsite generation will play a key role in servicing data centres in future, he suggests, not least because the major electricity providers – particularly in the US – are becoming increasingly concerned about their ability to meet demand.
An energy conservation drive could be bolstered by a greater use of fresh-air cooling, he adds. Today’s high-end computer equipment can happily operate within the external humidity and temperature ranges found for “nine to 11 months of the year,” says Newcombe. Instead of taking carefully-conditioned, filtered air from the blade chassis, passing it over the coolers and re-circulating it, “you can just take it from outside – and at least lose the chillers”.
CROP ROTATION
Perhaps more controversial, however, was Newcombe’s futuristic vision – what he dubs the “crop rotation” data centre model. “We have to break the mould, stop using the same old, tired, recipes for how we build data centres,” he says.
At the heart of the “crop rotation” notion is a continuous investment programme, that “continuously builds out capacity just in front of demand”, explains Newcombe. Over a four-year cycle, new modules are added, each using equipment that benefits from advances
in processing power and energy efficiency. At the start of the fourth year in the cycle, systems in the first module are readied to either migrate to new hardware or to be retired.
At this stage the model has yet to be road tested, but at the very least, it offers a challenge to today’s ingrained thinking, providing a model that “doesn’t try to predict long-term trends, allows businesses to respond quickly and effectively to change, all without the risk” associated with monolithic data centre construction, says Newcombe. Those qualities ensure that it won’t be untested for long.