As much of Canada suffers through a scorching summer and as increased power consumption strains hydro grids, the odds of a brownout or a blackout are increasing, and IT managers would be wise to get ready, just in case.
Hardware manufacturers are paying increasing attention to power consumption and releasing more power-efficient offerings, but aside from tearing out your infrastructure, there are some things you can do in the short term.
The first thing to do is make sure you’re ready for a power failure. Farah Saeed, program manager, backup power solutions with Frost & Sullivan, said the level of preparation will depend on the size of the data centre and how critical the business processes it supports are.
Today, though, she said most companies employ an uninterruptible power supply (UPS) and IT managers should ensure that those are fully functional. Saeed said many UPS manufacturers offer remote monitoring services to ensure UPS systems are functional and fully charged, removing the headache for IT managers.
Sometimes, though, the stress that a heat wave puts on power supplies can still reveal heretofore unknown issues. That’s what happened to Kern Weissman, director of infrastructure and network systems at CD&L Inc., a logistics and delivery company in South Hackensack, N.J. Power supply problems in New York last month affected a three-phase power supply at a secondary data centre in CD&L’s Manhattan office. But the UPS system didn’t immediately alert the IT staff to reduce voltage on one of the phases.
“The heat wave exposed an area of risk that I hadn’t considered,” said Weissman. There was no system downtime, but new automated alert procedures were quickly established, he said. Besides making sure you’re ready should the power fail, companies can also take steps to reduce their own power consumption and lessen the strain on the hydro system.
Tim Margeson, general manager of Toronto-based CBL Data Recovery Technologies, said that at the desktop level, not many companies have corporate policies in place to reduce power consumption. Even mandating that each desktop goes to sleep after 15 minutes of inactivity would help, he said, as would turning off unused monitors.
“They might want to look at some policies where users turn off their computers at the end of the day or when they go on vacation,” said Margeson. “A lot of places will just leave them on.”
Kenneth Brill, executive director of The Uptime Institute Inc. in Santa Fe, N.M., said another thing that data centre managers can do is investigate whether their cooling systems are delivering at their rated capacity.
He said half of the cooling units do not work efficiently for a variety of reasons, including clogged filters, belts slipping and refrigerant that’s not fully charged. If the air cooling system isn’t meeting a data centre’s needs, “the day they would discover that is on a hot day,” said Brill.
While more power-efficient servers are coming online, Kevin Smith, brand manager, servers at Dell Canada, said companies really need to look at their overall strategy with an eye to simplifying operations and increasing utilization, through technologies like virtualization.
The old paradigm was ‘One machine for every workload’, and Smith said the root cause of most data centre power problems is that overcapacity.
“If we just focus on reducing the power consumption of that piece of hardware, it’s almost like we’re taking a customers’ temperature when they have a fever but we’re not asking why they’re sick,” said Smith. “If we take a look at their environment, we just find there are too many servers for what they’re doing.”
He added that Dell has also developed best practices around data centre architecture and layout that can help companies reduce power consumption, by doing things as simple as rearranging servers to make more efficient use of the air conditioning system.
— with files from IDG News Service