The server farms which manage and route Internet traffic are among the most energy intensive venues on the planet. According to UC Berkeley researchers Justin Ma and Barath Raghavan, this use accounts for fully one percent of the 16 terawatts of energy used by humanity. That’s why researchers at Rutgers School of Engineering recently installed 16 polycrystalline solar panels on the roof of its computing center. The array, called Parasol – and comprised of up to 160 servers housed inside a stainless-steel-framed metal building the size of an SUV to protect the batteries, transformers and other electrical equipment – will be used to assess how to handle data and computer processing using solar-generated electricity.
Introduced at April’s EuroSys 2012 (11-13 April) in Bern, Switzerland, Parasol’s facility was largely completed in May under the direction of Rutgers Computer Science Professors Ricardo Bianchini and Thu D. Nguyen, with the assistance of post doctoral students and graduates.
As part of the funding for this project, the facility received more than $1.5 million in National Science Foundation grants given to Bianchini. Bianchini is also directing computer scientists from three other American universities (California, at Santa Barbara, Michigan and Virginia) using a three-year, $1.5 million grant from Google to assess energy use in big data centers.
The team aims to determine how to best use green energy via algorithmic simulations which take into consideration the amount of daylight, cloud cover, and seasonal changes in the angle of transit of the sun. Those simulations have been used to ascertain Parasol’s usefulness even before it was taken online, and resulted in the development of Green Hadoop and GreenSlot (a PDF), software programs that balance energy use by taking full advantage of clean, renewable energy. The team is now working on GreenNebula, an offshoot of OpenNebula, an open-source cloud computing toolkit probably not much different from the green EV program Toyota and Microsoft announced last year.