Berkeley lab looks to cloud for large scale energy savings

By Lou Fancher Correspondent Contra Costa Times

A six-month study by the Lawrence Berkeley National Laboratory at the request of Google has discovered a way to save 326 precious Petajoules. No, Petajoules are not a highly-endangered species you've never heard of--they're energy, as in electricity.

And in today's power source-hungry Internet world, where data centers account for 1 to 2 percent of global electricity use, a savings equivalent to 23 billion kilowatt hours (equivalent to the amount of electricity used annually by a city the size of Los Angeles), is significant. There's no magic wand: all it takes is moving three simple business applications from local computer systems from desktop to being hosted remotely via the cloud.

Working with Northwestern University's McCormick School of Engineering and Applied Science, LBNL principal investigator Lavanya Ramakrishnan and her team evaluated the effect of moving email, productivity software (word processing and spreadsheets), and customer relationship management software (file sharing) to cloud-based servers. To get the big numbers, they plugged best-estimated usage values representing the 86.7 million United States workers using these three most-common applications.

"We looked at the carbon footprint, the environmental benefits of moving these business services," said Ramakrishnan, in an interview after the study's findings were announced.

Although the results of moving high-use applications from under-optimized local servers to more efficient cloud servers did not surprise her, Ramakrishnan said the lack of available hard data did.

Some information on current businesses' energy usage was proprietary, and therefore unavailable, but mostly, it was nonexistent. "It struck me really hard: The energy efficiency measurements of servers and devices and usage was not there," she said.

That makes the model BNL used for the study all the more vital. Called CLEER (Cloud Energy and Emissions Research), a core component of the project was creating an open-source model for researchers, businesses, organizations and even consumers to use. The model is easily accessed online, meaning a person, IT department or data manager can view the study results, or input their own data in a choice of scenarios. The analysis tool can compare one business to another, show the energy and emissions impact of moving only select services to the cloud, run future cloud scenarios and even account for flex-factors, such as differing electrical grids. Results can be saved, downloaded, and made private under a user-approved login.

"This is the first study to do a rigorous, regional level examination and put out models that others can use," Ramakrishnan explained. "We envision a case where individuals and organizations can come with data targeted to plug into the model. It could be as small as a two-person company, or as large as a Fortune 500 company."

"We put the model out there for people to test drive," said LBNL Computing Sciences Communication Manager Jon Bashor. The model is not exclusive to Google, he emphasized, saying, "Google wanted an independent study and came to the lab because we have a long history of studying energy efficiency and environmental impact. We were doing it before it was popular."

"The thing I am most passionate about is the public use model," Ramakrishnan agreed. "I see a lot of scope if we get more data. We just opened the door and we're looking for feedback from all sectors."

Ramakrishnan is also looking at another Google-generated study: cloud-based video streaming. She believes the results LBNL will release from that study (now in draft stages) will be even more beneficial to individuals. But that thought doesn't stop her from offering a concrete example of how open source modeling for the latest study could impact real life.

"Imagine a large, dispersed school district where each computer is licensed to just one person. What would happen if the computers were licensed to one (cloud) center?" she asked.

The answer, energy savings, she claimed would come from efficient use of each computer, shared energy sources for an entire district, and the prospect of no more tech people driving from one school to the next for repairs.

"There's never a finish line for a researcher," she admitted. "It gets moved as soon as we get there."