The Energy Department is inviting researchers to apply for a space in its $32 million Magellan cloud computing evaluation program that's exploring the effectiveness of the cloud for a range of scientific computations.
Energy's National Energy Scientific Research Computing Center, operated by the Lawrence Berkeley National Laboratory, along with the Argonne National Laboratory, began setting up the Magellan project six months ago using economic stimulus law money, according to statements on the project Web site.
The program, based on IBM's iDataPlex server, offers 100 teraflops, or 100 trillion floating point operations per second, of total computational power. Scientific research in the cloud will include computations for astronomy and for genetic research.
Under the Magellan project, the cloud will be available to NERSC’s 3,000 scientific researchers, along with other selected researchers who have expressed an interest in the testbed. Staff members will use performance-monitoring software to analyze what kinds of science applications are being run on the system and how well they perform in a cloud.
One of the goals of Magellan is to examine how cloud computing compares with existing supercomputers. Many mid-range scientific projects do not require the petascale supercomputers and can be performed with small computing clusters.
Those types of mid-range studies may benefit from the Magellan cloud applications because it would eliminate the need to set up their own clusters with custom software. The cost of owning those clusters, along with costs for software, energy, space, security, operations and maintenance is motivating a look at other solutions, including cloud computing.
Under the program, NERSC and Argonne will create several computing environments in a cloud test bed. The project wants to determine:
Which scientific applications are best suited for the cloud.
The deployment and support problems with building large scientific clouds.
How commercial clouds can be leveraged for science.
How well does cloud computing support data-intensive scientific applications.
The security challenges for scientific clouds.
The projects also will deal with the problems of building a private cloud with a lack of software standards. Researchers will look at several products that create software frameworks for the Magellan cloud.