ORLANDO, FLA.--As if monitoring the Weather Channel for ice storms and hurricanes weren't enough, data center managers may have to track how space weather can affect their facilities. Eric Gallant, a data center consultant for design/build contractor Lee Technologies, said events beyond the earth's atmosphere – particularly on the sun – could have detrimental effects on the electrical grid and, therefore, on company data centers."Not all natural disasters originate here on earth," he said, speaking at the AFCOM Data Center World conference. "Some of them come from outer space." Gallant spoke primarily about solar flares, which are like explosions on the sun. Such solar events can cause major disruptions to the earth's magnetic fields which, in turn, could knock out the power grid for hours, or even days. Attendees took the presentation for what it was worth – prepare as much as they can, and probably as much as they would for a more earth-bound natural disaster. Make sure they have diesel generators and fuel storage. Run redundant systems. Test and maintain systems regularly. But building
Tim Richards, a facility manager at the Tennessee Processing Center for the Bank of New York Mellon, said he runs his data center at 78 degrees Fahrenheit. That's higher than normal, although still within recently expanded recommended guidelines from the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) for computer rooms. Richards said raising the data center temperatures saved the facility $250,000 in electrical costs. Asked about the risk-reward of hitting thermal limits quicker in the case of a mechanical cooling failure, Richards said there is so much redundancy that the company doesn't even worry about it. The facility's power usage effectiveness (PUE) – a comparison of total data center power consumption to IT power consumption, improved 30%, from 2.3 to 1.6, he said. And when asked whether the raised temperature created adverse working conditions, especially in the hot aisle, Richards said the bank generally runs a lights-out data center, so it isn't a major issue. Some data center stats and best practices
One early session on data center best practices led by Mark Levin, a senior partner at Metrics Based Assessments, highlighted familiar—and not-so-familiar stats, such as the following:
- Data center unit cost, measured in dollars per MIPS (a primarily mainframe measurement that stands for millions of instructions per second) or dollars per server image is decreasing
- Staff productivity, measured as full-time equivalents per MIPS/image, is improving.
- Server hardware, processor and disk costs continue to drop
- More people looking to build new data centers are considering colocation as an option
- Most data centers don't have the physical infrastructure to support new hardware technology
- Close-coupled cooling, often called in-row cooling, is gaining more traction just because of the increasing density needed for IT equipment
- Fewer raised floor data centers, coupled with more emphasis on ducted air return plenums in the hot aisle
- More emphasis from server manufacturers on developing products that can operate in higher temperatures
- More real-time monitoring of the data center, and in particular, the integration of facility information into the IT side.
Let us know what you think about the story; email Mark Fontecchio, News Writer. Also, check out our blogs: Data Center Facilities Pro, Mainframe Propeller Head, and Server Farming. ,/i>