| Insight

 
Browse ALL Categories   Browse ALL Categories
|
 Order #
Account *
 \
iQ Hidden Savings: The Imbalance of Power

iQ Hidden Savings - The Imbalance of Power


With energy prices on the rise, the green lobby gaining increasing momentum, and IT power consumption doing both, energy awareness and efficiency in the datacentre are no longer optional – and yet it needn’t be a chore, writes Christine Horton.

Imagine you’re running out of server capacity and there’s nothing else for it but to bite the bullet and invest in some new hardware. You would, in all likelihood, take reasonable care in obtaining the replacement kit. You’d think carefully about the exact requirement; do some research, look at the options, consider your budget, factor in business strategy – you know, shop around a bit.

What you certainly wouldn’t do is call some random supplier and have a conversation anything like the following:

“Oh yes hi. Can you please send me some servers?”

“No problem. OK, what are they for? How many do you want? How much do you want to spend? What brand, specifications, speeds, storage, CPUs, and so on would you like? When do you want them?”

“Oh, er, I don’t know. Just send me a selection. Whenever.”

“Er, OK. But how many?”

“Well, we need say four or five... better make it a couple of dozen just in case...”

But that’s just silly. Of course you wouldn’t do anything so ridiculous. What business would?

Well, perhaps the kind of business without the first clue as to how much power its technology department is consuming on a yearly basis – or, more specifically, a massive 89% of large UK companies according to a 2008 survey. Not convinced that the two scenarios are comparable? Well try this on for size.

According to IBM, over a three year period, the cost of running a server in a datacentre will outstrip the capital cost of the server itself. This, as well as other key financial and operational drivers, makes optimising datacentre power consumption a major priority, says Steve Yellen, VP of Product and Market Strategy at datacentre power management specialists, Aperture. The problem being that you can’t cut back if you don’t know how much you are using. It’s not like switching off the light when you leave the office.

“Today’s datacentre is an always-on facility, supporting the business round the clock. So it’s vital that it is as energy efficient as possible. Unfortunately (according to an Aperture survey) only 24% of datacentre managers measure their energy efficiency. That’s a pretty startling statistic. More than three quarters have no idea, and no easy way of finding out, how energy efficient their datacentres are. Because they’re not measuring, it’s impossible to benchmark performance or to track improvements.”

Another important factor, he says, is that even where companies are measuring their power consumption, they are often employing the wrong metrics. “There’s been a focus on improving the efficiency of the datacentre as a whole, as measured by ‘Power Usage Effectiveness’ (PUE), and not enough attention is being paid to optimising the computing infrastructure.

PUE is calculated by dividing the amount of power entering a datacentre by the amount actually being used to run the computer infrastructure within it. But you can achieve a great PUE ratio even if the computing infrastructure is inefficient, as long as you don’t provide too much more power than that infrastructure requires.” (In other words, as long as the datacentre is using all the power you throw at it your PUE will be fine... which is a bit like telling your kids that they can have as much pocket money as they like as long as they spend it all.)

With facilities and IT teams often sharing responsibility for the power bill, and no holistic management in place, traditional lines of departmental demarcation are a further issue, says Yellen. “Typically, facilities own everything up to the rack and the IT team picks it up from there. That makes it difficult to meet the challenge of cutting power bills.” He argues that IT and facilities teams must now collaborate more closely via a new metric that reflects both their individual and joint efforts to save power, and that scales from the device to the whole datacentre. This facilitates a ‘cascade’ effect, where changes at the device level translate into considerable savings at the facilities level.

According to Yellen, every watt saved at server component level (processor, memory, hard disk) results in an additional 1.84w saving in the power supply, power distribution, UPS, and cooling systems, and in the building entrance switchgear and medium voltage transformer. With this in mind, Emerson Network Power has proposed a metric of Compute Units Per Second Per Watt (CUPS/W), which takes account of improved IT performance as well as improvements in energy efficiency, to show how much computing power each watt is buying.

Looking carefully at where IT provisioning decisions are being made can also deliver some interesting insights. The No.1 cause of increasing power consumption in the datacentre is an increase in demand for IT services, says Yellen, so making line of business management accountable for the energy their applications consume can have a significant impact. It isn’t a strategy that has yet been widely adopted, but one approach is to charge individual departments for the energy consumed by the IT they commission. At a device level, the first step is to eliminate any obvious wastage and to optimise through consolidation, virtualisation and so on, which can be relatively quick wins.

After that, look at investing in more energy efficient devices. Specifics will depend on the installation and the precise IT requirement. “Some say that high-density equipment is more efficient because it means fewer devices and because a high density profile is more efficient to cool”, explains Yellen. “But it will be important to measure the impact of those devices on the entire datacentre, including the impact on the power required for cooling.”

“Increasingly, we expect to see companies moving towards utility computing so that they can be more flexible in responding to the business without having to ring-fence excessive capacity for applications. To reiterate, the key is to make savings early in the ‘electricity food chain’ to trigger the (1w = 1.84w) multiplier effect. Make tiny tweaks across the whole datacentre and optimise the floor space and cooling and you can unlock massive savings for the facility as a whole.”

Also, while it may lead to sighs of consternation from the gallery, it is impossible to have any discussion about power consumption these days without talking about the green issue, and this too has to be a consideration even for the hardest-nosed, most implacable IT manager. Indeed, although there is some cynicism about green technology – datacentre managers aren’t willing to trade reliability or availability for energy efficiency, and are reluctant to trust vendors’ green claims without convincing proof – many businesses are warming up to the fact that going green and cutting their energy bills amounts to the same thing.

“Despite the economic downturn we expect to see companies continuing or extending their green initiatives for that reason”, says Yellen. Franek Sodzawiczny, Development and Partner Director at Sentrum goes further still – citing the 'green initiative' in general as “quite possibly the trigger to re-ignite the world economy.”

“Datacentres are now taken for granted as a necessary business tool. (But) they consume 4% of the world’s power, and this will likely grow at pace as demands on data continue to increase.” In this regard at least, notes Sodzawiczny, the majority of businesses still have some way to go before they can really claim to be embracing green and financial drivers as one and the same concept – especially when it comes to their datacentres.

In recent research into attitudes towards green IT, just 20% of respondents thought the development of a mega datacentre would be in conflict with their organisation’s environmental policies. Data it seems – empowering it, accelerating it, disseminating it, sharing it, and protecting it – still, quite rightly, takes precedent over all other considerations. But one wonders for how much longer.

Streamlining your datacentre’s energy usage

Right-size your NCPI

Use modular, scalable power and cooling architectures to avoid “just in case” oversizing.

Virtualise your servers
Free up power and cooling capacity by consolidating applications into fewer physical (typically blade) servers.

Efficient floor layout
Floor layout can have a huge impact. Employ a hot-aisle/cold-aisle configuration with suitable air conditioner locations.

Efficient power equipment
A best-in-class UPS will significantly improve efficiency at typical operating loads. Remember that light-load efficiency is the key parameter, not full-load efficiency, and that UPS losses must also be cooled, doubling their cost.

Locate vented floor tiles correctly

Raised-floor datacentres often don’t have the right number of vented tiles; nor correctly located ones. Correct locations aren’t intuitively obvious so get a professional assessment.

Install energy-efficient lighting
Deploy more efficient lighting. Lighting power must also be cooled, doubling its cost. Turn off some or all via time-of-day or motion settings.

Utilise blanking panels

Blanking panels reduce hotspots and save energy by increasing the CRAC return air temperature. New snap-in toolless blanking panels make installation easy and inexpensive.

Source: APC

Airs and Graces

Just getting your air-conditioning right can have a massive impact on datacentre efficiency.

Row-based units (as opposed to room-based cooling) promote higher efficiency in high-density environments. Shorter air paths require less fan power. CRAC supply and return air temperatures are higher, preventing dehumidification and greatly reducing humidification costs.

Many air conditioners have economiser options. Use them.

Coordinate – multiple air conditioners may actually undermine one another’s performance. One may heat while another one cools; one may humidify while another one dehumidifies – resulting in massive waste.