Data center simulation technologies take the guesswork out of energy consumption

Researchers at Fujitsu Labs have developed technology that can predict energy consumption into today’s virtualized technology.

Among the many technologies I’ve covered, some of the most fascinating to me are the predictive types that determine what-ifs. For example, I’ve written about sensor and wireless technologies that can monitor water plants to keep tabs on just how pure the water is, and I’ve written about mesh networking to track environmental changes within bee hives.

So when I saw this news about Fujitsu Laboratories Ltd. and Fujitsu Laboratories of Europe Ltd. developing simulation technology that’s designed to help predict a data center’s energy consumption – particularly when something happens, such as shifting loads in a virtualized server environment (you know, the lofty cloud computing) – my interest was piqued.

As a backdrop, we all know that cooling is a critical element of any data center. But you can’t “try out” temperature fluctuations in an operational data center, even though cooling needs fluctuate depending on the time of day and season of the year, and in response to server utilization and server temperature. That’s why current cooling systems are designed to maintain a fixed room temperature. (And by the way, cooling is an energy-hungry element of a data center too – one that accounts for about 30% of a data center’s total power consumption, according to Fujitsu.)

That’s why Fujitsu decided to develop real-time simulation technology that could simulate today’s virtualized data centers, which have ever-changing operational profiles that make it more difficult to predict how changes to the operating status of hardware and air conditioning equipment impact the data center's overall power consumption.

The technology Fujitsu developed can determine thermal flows inside a building at calculation speeds over 1,000 times faster than was previously possible, as well as technology that completely models the hardware, air conditioning and other equipment of a running data center. When combined, the sensor and modeling technologies enable accurate calculations of a data center's power consumption and will make it possible to test the effectiveness of energy-saving measures before implementing them. The technology can instantly simulate the ways that power consumption is changed in accordance with server load concentration and air-conditioning controls.

Features of the newly developed technology, according to Fujitsu, are:

• High-speed thermal flow simulations which analyze the flow of heat within a building. The researchers reduced the number of required calculations by automatically extracting temperatures and flow patterns in advance, resulting in a thousand-fold increase in computing speed.

• By modeling all electrical and thermal flows, including hardware, air conditioning equipment, and power supplies, the simulation of the power consumption of an entire datacenter has a margin of error of less than 5%.

According to Fujitsu, the technology makes it possible to accurately simulate the effects that power-saving measures will have on a data center as a whole. For example, a data center operator could see what happens to power consumption when concentrating processing loads on servers in a fixed area during times of low demand, while at the same time cutting off power to other servers and reducing air conditioner output. The technology also makes it possible to take into consideration the climate conditions (temperature and humidity) at a data center's site and examine the cooling methods most suited to the location, Fujitsu says.

Now, the technology will undergo field testing and application at Fujitsu's data centers. Stay tuned…

What’s wrong? The new clean desk test
Join the discussion
Be the first to comment on this article. Our Commenting Policies