HP Sells Heat Modeling Service to Cool Data Centers

The real problem with data centers is that they are on CRAC. No, that is not a twist on the Gaelic word for romping good fun, craic, but rather the short-hand for computer room air conditioning unit. This is the device that cools the air, removes water from it, and pumps cold air into parts of a modern data center, thereby keeping servers, storage, and networking equipment from melting. CRAC units also take away heat, which gets sucked into HVAC systems and vented to the outside world. If Hewlett-Packard is right, companies are not necessarily using CRAC correctly, and it has a service to fix that.

This week, HP announced a service called Thermal Zone Mapping, which basically puts monitors in the raised floor and dropped ceilings of a data center as well as every two feet up a rack of servers to build a three-dimensional view of heat flows and cold flows in the data center as workloads are running on racks of servers. According to Brian Brouillette, a vice president in the company’s HP Services unit, which is selling the TZM product, HP Labs took a bunch of PhDs and some computational fluid dynamics software and created a system to analyze what is going on with the CRAC units as they fight heat in the data center. The TZM software can capture data on a daily, weekly, and monthly basis and show companies exactly what is happening in the data center–it is a bit like modeling the weather inside the room, in fact. If you have ever been in a data center, there are hot and cold spots, and getting the right amount of cold to cancel out the heat islands where servers or storage are running is a tricky business. Which is why HP thinks it can charge for a service that helps companies who are trying to pack more gear into data centers without blowing fuses.

Brouillette says that a lot of HP’s larger customers have run out of data center capacity because they can’t deploy new servers and storage since they are out of power for both the units themselves and the CRAC units to cool the room where they sit. In some countries, like China and India, getting power is a problem in and of itself, so minimizing electricity use is a starting point in designing a new data center. And even in Western economies, power is becoming an issue. For instance, the city of Palo Alto told HP itself some time ago that it would have to put a cap on the electricity it uses, or else its growth would cause brownouts in the city.

This is one reason why HP is undergoing a massive data center consolidation effort over the next three years, moving to three redundant data centers from the current 85 data centers the company runs worldwide. The TZM service is one of the ways HP is able to do this consolidation, and so is a set of companion technologies called Dynamic Smart Cooling, which is comprised of a few different technologies. DSC includes electronics in HP’s server and storage products to throttle back the amount of energy they use and therefore the heat they generate; it also includes air-conditioning products from various HP partners, including Liebert and Shultz, to dynamically direct cooling to hot spots in the data center. HP is using TZM and DSC in its Palo Alto and Austin data centers (the other one is in Houston), and it projecting that it can cut energy usage in the data centers by 10 percent compared to not using them, and helps get HP on track to reduce its global energy usage by 20 percent by 2010.

While HP is expecting big results in its own data centers, Brouillette says that customers should expect anywhere from 10 percent to 45 percent improvement after HP’s experts and partners come in. The TZM service, which has a starting price of around $100,000, scales according to the square footage of floor space in the data center.

That may seem like a lot in a world where a pretty powerful server costs only $2,500, but an early customer experience shows how desperate some data centers are to add computers but not use more power and cooling–mainly because the average new data center costs somewhere around $10 million. One anonymous customer in London in the financial services business had 4,000 blade servers deployed, and it needed to add another 1,000 but could not. After HP came in with the TZM service, the CRACs were tuned and the gear moved around in such a way that another 1,000 blade servers could be added to the data center and still stay in the same power envelope. That same company is now looking at DSC technologies to see how many more servers can be squeezed in the data center.

RELATED STORIES

Dell, IBM Push Power-Saving Servers

The Balance of Server Powers

                     Post this story to del.icio.us
               Post this story to Digg
    Post this story to Slashdot