Microsoft is giving scientists and researchers supported by the National Science Foundation free access to its Azure cloud computing platform for three years. The deal would potentially, in one fell swoop, give Microsoft access to research projects that the NSF funds in nearly 2,000 universities and institutions in all 50 states.
The agreement would also give Microsoft access to computer scientists who could help the software giant to develop new ways to simplify the use of cloud computing for the scientific and research community.
Microsoft formed the eXtreme Computing Group in June 2009 as part of its Microsoft Research arm. Its goal is to develop “radical new approaches to ultrascale and high-performance computing hardware and software,” specifically for computer security, cloud computing, and data center architecture, among other areas.
“I am very excited about this, as it is the fruit of nearly two years of planning and collaboration across Microsoft product and research teams, as well as many discussions with researchers, university leaders and government agencies,” wrote Dan Reed, corporate vice president for Microsoft’s eXtreme Computing Group, in his blog.
‘More Than Adequate’ Horsepower
Microsoft didn’t detail how much Azure capacity it would give away, but during a conference call announcing the deal, Reed said the compute power would be “more than adequate” for researchers’ needs.
Jeannette Wing, NSF assistant director in the Directorate for Computer and Information Science and Engineering, said the NSF would invest $5 million into this agreement, and is inviting scientists to compete for the opportunity to use Azure. “We’re encouraging computer scientists to propose to us innovative uses of the cloud to enable general scientists to make new discoveries,” she said.
Reed said Microsoft is challenging computer scientists to create ways to simplify the use of cloud computing for the scientific research community. He said it should be as easy for researchers to use cloud computing to mine vast amounts of research data using familiar applications, such as Microsoft Excel.
Cloud-Based HPC ‘Underexplored’
Wing described the use of cloud computing by researchers as “underexplored”, though it is something that cloud computing providers are pushing. Last year, HPC Linux cluster provider Penguin Computing launched Penguin on Demand (POD) a cloud offering to the HPC community, while Amazon is wooing educators, academic researchers, and students with free usage credits for Amazon Web Services.
Microsoft and Amazon have the marketing dollars to offer free resources to the research community that often have budget restraints. But Penguin Computing CEO Charles Wuischpard says general-purpose clouds are not suitable for HPC applications.
“In a cloud environment, there is no guarantee where the virtual machines will run, their proximity to each other, or how close they are to storage,” Wuischpard said. “POD is a highly optimized HPC cluster that runs jobs on physical compute core that are in the same processor, in the same system or tightly coupled through an InfiniBand interconnect. POD compute nodes are directly attached to storage on a 10GigE network – both standard NFS and high-speed parallel file system storage is available (at the user’s choice). IO is critical to the performance of many HPC user applications – and IO really suffers in a decentralized, virtualized environment.”
“Cloud computing environments don’t scale out reliably to the size that HPC users require. Launching 60,000 jobs that consume 100% of 640 cores is not uncommon,” Wuischpard added.
Microsoft said it will offer support to the scientific community. As Reed wrote in his blog: “… a technical computing engagement team, led by Dennis Gannon and Roger Barga (principal architect, Cloud Computing Futures at Microsoft Research), will work directly with NSF-funded researchers to port, extend and enhance client tools for data analysis and modeling.”
Azure joins other cloud technologies to which NSF scientists have access. Among those include a set of cloud-based software services supported by Google and IBM, and a cluster managed by HP, Intel and Yahoo housed at the University of Illinois at Urbana-Champaign.