Supercomputers consume a huge amount of water to cool down servers through towers that are typically located on the roof of supercomputer facilities. For example, a 15 MW data center could use up to 360,000 gallons of water a day, according to Amazon's James Hamilton. However, due to recent drought conditions around the United States, there have been concerns over the tremendous amount of water required to run data centers and supercomputers.
Facebook and eBay have developed dashboards to monitor the water efficiency in run-time, while Google and the NCAR-Wyoming Supercomputing Center are developing water-efficient cooling technologies, such as using outside air cooling and recycled water. However, Florida International University (FIU) professor Shaolei Ren notes that these approaches only aim for facility or infrastructure improvement, and they require high upfront capital investment and/or the right climate conditions.
Ren says FIU researchers are working on software-based approaches to mitigate water consumption. The researchers found that the spatio-temporal variation of water efficiency also is a perfect fit for supercomputers' workload flexibility. During a demonstration of their approach, the researchers reduced water consumption by 20 percent with almost no compromise in aspects such as service latency.
From HPC Wire
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found