The year 2008 represents a watershed for corporate IT. While the cost of running the data centres that power most businesses has historically been dominated by the price of the IT equipment they house, from 2008 the cost of the energy required to power that kit will – for the first time – surpass the cost of the computer kit itself.

The soaring cost of energy has, of course, accelerated that change. But a more significant factor is the rapid rise in the amount of power being drawn by today’s servers, networking equipment and storage systems, and the equally large quantity of electricity being used to keep that from overheating.

 “Customers are facing the most difficult time in their career,” says Aaron Davies, chief marketing officer at data centre infrastructure management company APC.

“There are power and cooling problems right across the IT landscape. The fastest moving industry in human history has outpaced the infrastructure’s ability to keep up.”

There has been a dramatic increase in the appreciation of the seriousness of that situation over the last six months, says John Coulson, enterprise marketing manager at Dell in the UK.

Cool runnings

As organisations invest in IT infrastructure they are becoming acutely aware that not only do they have the energy demands that come with their new products but the energy demands of cooling that environment.

“It is not one product equals one watt, it is one product equals the watt associated with that product plus the wattage associated with cooling that products,” says Coulson.

APC’s Davies echoes that perception: “The one thing the IT community is beginning to understand is that power in equals heat out, equals cooling problems.”

That means that some IT decision-makers have bought servers they can’t turn on for a lack of power. It means that people are buying equipment and bringing it online, only to find that they don’t have enough cooling to keep it operational.

And that can be a career-threatening situation. Davies tells the story of one CIO who invited the company’s senior management team to the ribbon-cutting ceremony of his new data centre. But two days before the ceremony, he found he could not turn the facility on because of a lack of power.

And the dearth of adequate power will only get worse.  Davies predicts that more than half of all data centres are going to have to move location in the next few years because they can’t get access to enough power.

Racked up

R&D teams at all the major infrastructure makers are working hard on the problem, aware that vendors who can reduce the amount of energy their equipment uses to process a given IT job will gain a significant advantage. But as energy-efficient as those can be made, there is little chance that tomorrow’s systems will require less power.

In fact, organisations that only a few years ago were drawing 1 to 2 kilowatts (kW) per rack are now looking towards 12kW a rack, and expecting a requirement for 20kW and above in the years ahead.

Indeed, many companies today are forced to run their racks half empty because they are unable to deal with the heat that would be generated by a fully loaded tower.

The challenge of dealing with that heat – and dealing with it in the most efficient way possible – boils down to optimising processes and technologies for removing it from the data centre. In other words, the proportion of energy used to remove the heat generated by processing equipment needs to be a much smaller fraction of the overall energy bill.

At present that proportion is unacceptably high. According to Neil Rasmussen, chief technology officer and co-founder of APC, between 30% to 60% of the energy used by most data centres is employed within the processing workload. Almost all of the rest is sucked up by cooling and power conversion.

The difference in the amount of energy relates directly to the type of data centre in question – if it is a bank or a payroll processing centre with lots of redundant kit to ensure 24x7 availability, then it is going to have a much higher burn-rate than a university data centre which might power down some kit during off-peak periods.

The upshot for any data centre though is that power and cooling issues are going to be a barrier to investment in infrastructure. “If you are spending half of your IT budget on energy, you can no longer buy all the new routers or the new servers or the hotshot web developers you could previously,” says APC’s Davies.

IT analysts at Gartner estimate that 48% of IT budgets are going to go on energy. That is up from a mere 8% in the 1990s.

“If IT executives thought they were facing a reduction in their budgets of 20% to 30%,” says APC’s Davies, “there would be mass hysteria right now.”

“Energy costs are going to be a hot potato; right now they are spread right across the company, but over the next few years [finance people] are going to put a ring around it and throw it back to the CIO, saying ‘It’s your problem, bring your consumption down’.”

Cold Cut

Cutting the energy draw of cooling equipment would therefore have a huge impact on both the energy bill of the data centre and (potentially) its overall carbon footprint. And although future cooling technologies promise a great deal, companies don’t have to wait for new equipment to emerge from the labs.

Taking 25% off of the energy profile of cooling processes is feasible today, argues APC’s Rasmussen.

In-row cooling is one key area. By cooling the equipment directly within the rows of servers, data centre managers can gain huge efficiencies over traditional technologies which are set up to cool the computer room as a whole. According to Gartner, by 2011 in-row and in-rack cooling will have emerged as the dominant cooling approach.

The seemingly simple approach of upgrading old UPS (uninterruptible power supplies) to the current generation of much more efficient systems is another approach that has the potential to cut in half the loss of power from UPSs.

The scaling of cooling equipment to the job in hand is also a key tactic: many companies oversize their cooling, and waste energy as a result.

Talent squeeze

Lastly, management software is going to be critical. According to a survey by the data centre managers group Afcom, the talent pool of those with data centre skills is shrinking dramatically. Over the next few years, almost 50% of the people who have experience of running data centres are going to retire, and there is no pipeline in place to fill their shoes.

That means adoption of management systems that monitor and balance power – which barely exist today – will become the norm, says Gartner.

There are other techniques – some proven, some being explored – hot-aisle, cold-aisle separation; drawing air from outside the building; pushing the temperature  up within the data centre to around 29C, the wider use of liquid cooling. The only option that is not on the table is inactivity.

“If you do things the same old way for the next few years you’re dead because you will automatically be running an inefficient system. You need to find a new way to get both availability and efficiency,” says Davies. And no one is saying that is going to be easy.