Everyone is after a cheaper data centre these days, and Hewlett-Packard Co. — along with IBM and Cisco — is one of the companies after those precious data centre budget dollars. Under the banner of HP Critical Facilities Services, delivered by EYP MCF (its 2007 acquisition), the company is fighting the battle against the bulging data centre.
“Eighty-five per cent of data centres built before 2001 are now obsolete,” said HP senior energy consultant Alda Licis at an event on Thursday. “Reliability and uptime demands have increased, and the cost of servers and rate of technology refreshes means a higher density in data centres.”
One of the biggies when it comes to pinching pennies in the data centre is new cooling methods. One good way to bring down those costs is to construct the data centre in a climate with “more free cooling days,” Licis said.
This means having the data centre in a location where a chillier climate means being able to turn down (or off) the air-conditioning units and save on cooling bills.
“Does your data centre have to be where you are? No,” said HP senior principal and vice-president of critical facilities assurance at EYP MCF Rick Sawyer.
Aaron Hay, research consultant with Info-Tech Research Group, said, “I think that overcooling — particularly in old data centres — is really common.” Another strategy, he said, is to consider outsourcing data centre operations to a hosting company that already has a data centre in a cooler area.
(It is important, said Hay, to do a full cost analysis when it comes to deciding between building a new data centre and porting data centre operations to another company, since the long-term rental can end up being just as pricy as the up-front costs of a new facility.)
Another trick is to actually raise the temperature in the data centre. “Servers can actually take higher temperatures and a wider range of humidity now. There’s been a lot of changes in the industry,” she said. “Before, when you walked into a data centre and it felt like a meat locker, you’d say, ‘Ahhh, everything’s working.’ Now, when you walk into a data centre and it feels like a meat locker, you think, ‘This could use some improvements.’”
Licis suggests raising the set-point of the temperature in the data centre by a degree or so at a time, and then taking measurements of how that exactly that affects the servers, as well as any energy savings.
“There really are very few side-effects (to raising the temperature set-point),” said Hay. “It’s a really quick move to take a few percentage points off your energy bill.”
Another important strategy is to ensure a minimum of downtime. According to a 2007 Data Centre Journal study, every hour of downtime in a big data centre costs the company around $1 million. “The single largest cause of these mistakes are human error, including bad design, bad operation, or bad personnel management,” said Sawyer.
That is why it is important to do probability risk assessments and benchmarking maintenance and operational practices to find what works, Sawyer said. “And don’t forget to bring in utilities for this,” he said.