Left to right: Michael Kirwan, CIO at Yahoo!, Rosio Alvarez, CIO at LBNL and Aditya Fotedar, CIO at Wyse Technology speak at the Silicon Valley Leadership Group's Data Center Energy Efficiency Summit in October 2009

From studying data center energy management to dedicating resources to exploration of cloud computing, the US Department of Energy’s Lawrence Berkeley National Laboratory in Berkeley, Calif., is a key institution for the data center industry, constantly churning out vendor-neutral information.

The lab’s IT infrastructure provides for a chunk of DOE’s computational needs and those of a multitude of scientists. Demand for the lab’s compute capacity is constantly growing and its data centers are running out of space, power and cooling capacity.

“We’re at capacity,” LBNL CIO Rosio Alvarez said in an interview. “We can grow a little more but we’re very close to capacity.”

Alvarez is responsible for ensuring the laboratory has sufficient IT infrastructure to support the work done there. According to a pre-panel introduction at the Silicon Valley Leadership Group’s Data Center Energy Efficiency Summit earlier this month, LBNL researchers carry out sponsored research funded by more than $0.7 billion in quantitative biology, nano science, new energy systems, environmental solutions and the use of integrated computing as a tool for discovery.

The division Alvarez oversees also runs infrastructure that supports major DOE user facilities, including one of its new nano science centers, a super computer center, an advanced light source and a multi-lab genomics institute.

LBNL’s main data center houses about 5,000 square feet of IT-equipment floor. Combined with several other small data centers, the laboratory’s total floor space measures about 10,000 square feet. All of the facilities are in Berkeley, Calif.

The lab addressed some of its capacity shortage by using Amazon’s Elastic Compute Cloud services. Some of the lab’s applications have been running in the EC2 cloud since spring, Alvarez said.

Cloud computing, however, does not completely solve the problem, as many scientific applications demand quicker speeds than the cloud is able to provide.

“The (workloads) that are not very good for the cloud are the ones that are very latency-sensitive,” Alvarez said. “They are the ones that are called MPI (Message Passing Interface) applications. They are … very tightly coupled so the latency between wherever the different nodes are in the cloud is too much for these applications.

“They won’t fail but they’ll run terribly slow – sometimes 10 times slower than in a regular bare-metal box.”

Alvarez’s team is under pressure to expand capacity because of growing demand and also because the lab is receiving a lot of the American Recovery and Reinvestment Act (stimulus) funds, which can only be used for projects that can be completed immediately.

“We can’t wait to build a new data center or retrofit (the existing one) some more,” she said. “Whatever it is, it has to happen immediately.”

That is why LBNL is now considering containerized data center solutions. As of mid-October, the organization had not settled on a vendor.

Related news: Scientists expose attack vulnerability within public clouds
Related feature: Lessons learned from one of Amazon’s cloud computing client’s prolonged downtime
Related feature: Container market diversity increases