NSF plans grants for researchers to use cloud platforms

The volumes of data being generated by today’s big science projects have outstripped the capacity of desktop computers, requiring computations that can take months to complete. With time on supercomputers being a limited resource, some researchers are looking to cloud computing as a way to access additional computational resources.

The National Science Foundation has partnered with a number of commercial cloud and cluster service providers to give researchers access to these resources, and to help NSF learn how scientific research can best take advantage of an emerging computing environment. As those agreements fall into place, including one with Microsoft announced earlier this month, NSF officials are beginning to contemplate how to make the best use of them.

“We are trying to understand what types of applications are most suited for a certain type of infrastructure,” said Manish Parashar, program director in NSF’s Office of Cyber Infrastructure.

Related coverage

Feds could get access to advanced cloud computing technologies

HHS plans move to cloud for regional centers

Under the Microsoft agreement, Microsoft will provide free computing time on its Windows Azure cloud computing platform to researchers NSF selects. The program complements NSF collaborations with cloud and cluster providers over the past three years. It will let researchers extend the power of desktop applications by moving them into the cloud where processing power can be accessed on demand.

Although desktop computers often are not adequate for analysis of the large data sets being produced by large science projects, supercomputers can be overkill. Much of the work could be done through parallel processing on multiple clustered computers. But assembling clusters locally to support projects can be expensive and inefficient. Cloud computing, NSF hopes, could be the solution to making centrally managed resources available on demand without the overhead of acquiring and managing the clusters for each project or working group.

But cloud computing is an emerging technology that still is maturing. Despite is promise, a lot of work must be done to understand how it could best be used for large-scale scientific research, Parashar said. To answer questions of how researchers can best use this environment, NSF has collaborated with a number of companies to provide access to commercial cloud computing resources.

In fiscal 2008 the NSF Computer and Information Science and Engineering office created the Cluster Exploratory Program, which gave access to a set of cloud-based software services supported by Google and IBM, and access to another cluster supported by HP, Intel and Yahoo housed at the University of Illinois at Urbana Champaign. These clusters support the open-source Hadoop programming interface. The Microsoft platform supports a Windows interface, giving researchers a new software environment to work in.

In addition to these programs, NSF’s Office of Cyber Infrastructure has funded the TeraGrid, an effort to provide high-end resources through a system of distributed platforms that includes a pool of computers at Purdue University. OCI also has funded the Gordon system, a platform for data-intensive computation at the San Diego Supercomputing Center, and FutureGrid, an experimental platform at Indiana University to conduct research in grid and cloud computing.

Microsoft Azure is a cloud platform that allows applications to be hosted and run at Microsoft data centers, using the Azure operating system. Research projects using the platform will develop their own applications for hosting, with the assistance of a Microsoft support team, to help researchers integrate their software into the platform.

The Azure program will be open to research projects selected and funded by NSF. Proposals can be submitted to supplement existing grants, or new grants can be sought under the Early Concept Grants for Exploratory Research program, a program to support high-risk research. EAGER grants are limited to $300,000 for a maximum of two years, but NSF expects to announce a new longer term grant program for Computing in the Cloud shortly, Parashar said.

To be eligible for the program's supplemental grants, the existing grant award must still be open at the time the supplement is awarded. Proposals for EAGER grants can be submitted by April 15 for consideration this fiscal year. NSF is asking researchers to be creative in the development of innovative applications that can challenge and exploit the new resources, and is encouraging collaboration between computer scientists and other scientists and engineers

“It is a little early for concrete proposals,” Parashar said. “But there seems to be a lot of interest.”

For additional information, contact Parashar at mparasha@nsf.gov, Chita Das at cdas@nsf.gov, Krishna Kant at kkant@nsf.gov, or Frank Olken at folken@nsf.gov.