The loss of power from the source of its production to where it is consumed in data centres is inevitable. However, there are ways to minimise it. Senior executives from both sides of the DC industry discussed how power leakage can be tamed at a roundtable discussion in Auckland.

Data centres have a somewhat patchy reputation when it comes to power consumption and greenness. As huge consumers of power, they are seen as a necessary evil by some. Still, they have the potential to serve as examples of how energy can be used very efficiently.

“I see huge opportunity for data centers, not only in enabling technology and delivering it to create a better society but also, as large power users, showing the way for others on what can be done,” said panel member Lee Stewart, the owner of ESG Strategy, which helps organisations align their environmental, social and governance goals with your current business strategy.

This exemplary role puts a greater onus on DC operators to make the fullest use of the power entering their facilities. Within DCs, power losses occur in several areas, including non-intelligent PDUs, cabling and connectors, transformers and UPS systems. The use of smart, high-efficiency types of these assets can help minimise power leakage while ensuring the power demand is as close as possible to the actual requirements.

Power is lost en route to a DC, too, mostly in the form of heat from the transmission grid – the longer the route, the greater the loss. To keep such leakage to a minimum, some enterprises are choosing to site their facilities near the sources of power production. One such company is T4 Group, a colocation provider that is currently building a Tier 4 carbon-neutral DC in Invercargill in South Island, New Zealand.  To be run using just hydropower from nearby plants, with 100 percent natural cooling, the modular facility is sited next to a hotel, which will use waste heat conveyed from the DC for heating.

“The main requirement for power supply is reliability and, even more so now, being green. Clients are actually interested in actually how much carbon you're actually using or producing, your carbon problem, and what you’re doing to offset it,” said Jason Porter, the company’s Director of Technology. “So it's become an important part of actually building a DC. There's no point in building a high-end DC and losing percentage points of the actual energy through leakage.”

Grid reliability is especially important in New Zealand, added Tamas Balogh, Data Centre Program Director, BW Digital as the grid is a local one with no international connection, unlike in continental Europe where a multi-country synchronous grid supplies electricity to over 400 million customers in 24 countries.

The company is building a 150MW hyperscale DC in Invercargill that will provide scalable hosting solutions and colocation services; the first 10MW module is expected to be completed next year. In the future, BW Digital hopes to become part of a local smart grid campus in which companies don’t just consume power but also contribute to it using onsite energy storage systems, said Mr Balogh.

Once new DCs are up and running, what will soon come into play is the ability to scale up quickly, said Sanjay Motwani, Vice President, APAC, Legrand. “The market demands it. It's not like you've got a two- or three-megawatt DC and it's going to take you through for ten to fifteen years; you would be in a situation a few years down the line. You need to scale up and you need to scale up quickly. So on the power supply side, you have to ensure that there is proper transfer merging and so on.”

To make the most use of power, it is necessary to measure it up to the device level, he added. While a good number of DC operators have now started measuring power at the rack level, those that do so at the device level form a minority. “You can easily get information on the stranded power within a rack if you measure at the device level because you will know how each device is behaving. Excluding hyperscalers, there would typically be about nine to ten ghost servers, idle servers which are consuming power. And if you're not tracking which server is consuming how much power, you're actually wasting power,” he said.

In agreement, Mr Porter commented, “End users may not care about their power, but as a DC operator, I do because if they're using more power than what they should, that means I can't give it to something else. So it's important that we actually do track at every level, at every point. And for security reasons, such measurement must be done in-house.”

The roundtable closed with the panel members all agreeing that no DC today operates in isolation. Rather, each operates within an ecosystem, one that is itself part of a bigger community. For DC operators, one new area that will need attention in the near future is the ability to store energy and giving it back to the grid. For now, the challenge is operating their facilities at very high efficiencies while consuming the least possible power.

The roundtable was moderated by Nick Parfitt, Lead Analyst for DatacenterDynamics publication in the Asia Pacific.