Jump to content United States-English
HP.com Home Products and Services Support and Drivers Solutions How to Buy
» Contact HP

HP.com home

Getting smart about data center cooling

New technology saves power, cuts costs


HP Labs

» Research
» News and events
» Technical reports
» About HP Labs
» Careers @ HP Labs
» Worldwide sites
» Downloads
Content starts here
Overcooling a whole data center just to prevent a few hot spots in a few key servers from failing is hideously inefficient. It's like using a chain saw to cut butter.

By Lisa Stapleton, November 2006

When HP Labs Fellow Chandrakant Patel started looking into energy conservation in data centers, awareness of the importance of IT power conservation was in its infancy.

"Five years ago, people didn’t get fired for wasting energy, but they could get fired if excess heat started to make servers fail," says Patel. "So they set their air conditioners lower than they needed to, and they left a lot of empty space in the racks to spread out the heat load. Power was cheap then, so that was a solution."

Today, it’s a different world. With energy prices ever-increasing, many IT managers are taking a more integrated approach to reducing costs. They’re recognizing that unresolved power and cooling issues can limit the use of all the "real estate" in their racks, because it’s tough to cool a full rack. And they’re looking for options.

Patel and his team are working to provide options -- products and services that address the new, brutal economics of running a data center. Today, lowering the cost of ownership is about getting the right power and cooling, to the right places, at the right times.

Problems multiplied

The problem is amplified because server needs have exploded. "Where businesses used to have a few mainframes, they now have 4,000 to 10,000 servers to support," says John Humphreys, program director for IDC’s enterprise platforms group. "The total number of servers has grown from about 6 million 10 years ago to 24 million now."

This large installed base is changing the cost of managing servers, says Humphreys. "Ten years ago, 70 percent of the total cost of acquisition and management of servers was in the cost of the hardware. Today about 70 percent of the cost of managing servers revolves around keeping the lights on -- power and cooling, maintenance, upgrades, for example -- and only 20 to 25 percent consists of the cost of the hardware."

Many managers resort to practices that they know cost them a lot. In particular, many are tired of paying to generate wasted heat, and again to remove it. A recent study by HP and the Uptime Institute suggests that for most of the world’s data centers, 63 percent of the power is associated with cooling the IT equipment.1,2

Many are also tired of overcooling a whole data center just to prevent a few hot spots in a few key servers from failing. Patel agrees this strategy is hideously inefficient. "It’s like using a chain saw to cut butter," he says.

Containing power costs

He knew there had to be a better way. HP Labs research and HP product teams have given power and cooling solutions a high priority, garnering numerous patents in the area. Recently announced power and cooling innovations -- including Dynamic Smart Cooling -- incorporate some of the key ideas of this cutting-edge research.

HP’s experience and research in data center power management means that customers can upgrade their servers and still contain power costs.

Mark Donahoo, senior infrastructure support manager at Apollo Group, says he was pleasantly surprised to find out that he could pack servers more densely without taking a heating hit. He replaced 32 standard, p-Class nodes with 48 new HP BladeSystem servers. "That’s an increase of 16 nodes, adding two rack units (U) to the rack, without additional power or cooling to the rack."

How dynamic smart cooling works

Dynamic Smart Cooling combines sensors with control nodes that monitor temperatures in parts of the data center. These nodes control the computer room air conditioners for better control of the whole IT environment, so you don’t have to cool a whole room for a few racks in one area. For example, using HP’s Dynamic Smart Cooling, HP Labs reduced the power to cool a data center by anywhere from 30 to 60 percent, depending on facility infrastructure.

Similar results recently prompted Peter Gross, chief executive officer and chief technology officer of EYP Mission Critical Facilities, to call Dynamic Smart Cooling "the most remarkable development for data-center-critical support systems."

This localized approach to controlling energy is being applied across HP’s server portfolio and other data-center products. Apollo’s Donahoo says he likes the control he gets with HP BladeSystem products. Particularly interesting, he says, are "the centralized cooling and power and the detailed Thermal Logic power settings, which include the ability to oversubscribe power and set power settings or limits, to allow a specific wattage to be allocated to a blade."

Infrastructure for power management

HP Modular Cooling System
» Buy or get info

In November 2006, HP also announced plans to bring to market a modular, three-phase power protection and distribution infrastructure designed to provide higher levels of power density to rows of servers and storage. Like HP’s popular Modular Cooling System, the new infrastructure takes an IT approach -- modular, redundant, manageable -- and applies it to common facilities challenges.

Patel and other thermal experts say the problems of power and cooling will yield if attacked on a variety of fronts. One promising idea is controlling power and cooling by using software that is aware of the locations and heat profiles of different processors in relation to the cooling resources in the data center.

The software would send different computing tasks to each processor based on temperature and the amount of heat. In other words, processors would be turned on and off, or given shorter or longer tasks, based on policies that balance power and cooling resources and needs. Similarly, servers -- both virtual and real -- could be assigned work based on sensing the temperature around them, and using thermo-fluids policies to achieve a balance of power and cooling i.e. result in the data center cooling resources operating at an efficient operating point.

Some of these ideas have yet to be incorporated into servers and data center management system. But the economics of data centers are fundamentally changing. Patel notes that reducing the power demands has an added bonus: Generating less power means consuming fewer natural resources and promoting a cleaner environment.

By making power and cooling a priority, researchers like Patel hope that, one day, the chainsaw-to-cut-butter approach will be a memory, replaced by a new slogan: "The right power and cooling for the job, at the right time, in the right place. No more, no less."

1Christopher Malone, PhD, Christian Belady, P.E., "Metrics to Characterize Data Center & IT Equipment Energy Use," Digital Power Forum, Richardson, TX (September, 2006).

Related links

» Smart cooling research
» Energy-efficient computing
» Dynamic Smart Cooling
» Press kit
» Special report: Power & cooling
» HP power & cooling technologies

News and events

» Recent news stories
» Archived news stories

Recent technical reports

» Model-Based Approach for Optimizing a Data Center Centralized Cooling System
» Energy Flow in the Information Technology Stack
» Local Temperature Control in Data Center Cooling: Part II, Statistical Analysis
» Cost Model for Planning, Development and Operation of a Data Center
» More publications
Printable version
Privacy statement Using this site means you accept its terms Feedback to HP Labs
© 2009 Hewlett-Packard Development Company, L.P.