Bridging the GAP
Bridges Will Bring High-Performance Computing and Data Analysis Capabilities to New Fields
Today, investigators in important, data-intensive fields such as cancer genomics, the digital humanities and machine learning increasingly find that they need the data handling and analytic power of high-performance computers operating in a system designed specifically for manipulating and storing very large amounts of data.
With its new Bridges environment, PSC will provide these researchers with a new level of computational and data-handling capability that meets with their needs. Supported by a $9.65-million National Science Foundation (NSF) award, Bridges will be delivered by Hewlett-Packard (HP) based on an architecture designed by PSC. Construction is scheduled to begin in October 2015, with a target production date of January 2016.
“We designed Bridges to benefit new communities and to bring the power of high-performance computing to Big Data. The community’s response to learning about Bridges has been overwhelmingly positive, confirming Bridges’ transformational potential for research.”
—Nick Nystrom, Bridges Principal Investigator and Project Leader, PSC
BRIDGING TO NEW RESEARCH COMMUNITIES
To support data analytics and to help research communities that have not traditionally used the tools of conventional high-performance computing (HPC), Bridges will offer features for flexibility and ease of use normally associated with lower-powered systems:
• extensive interactivity to let users work as
they do on their own personal computing devices, complementing thought and providing immediate feedback, rather than requiring users to submit jobs and await their completion as in the usual supercomputing environment.
• web-browser-based gateways that will launch jobs and manage workflows transparently, behind-the-scenes, on users’ behalf. This will let them harness Bridges’ capabilities without having to master the more arcane tools of HPC.
• users’ software and tools such as the Hadoop ecosystem, Python, R, MATLAB® and Java. Persistent databases and web servers will enable modern, flexible, and easily extensible software architectures.
• software portability and reproducibility, greater user control over software and environments and cloud interoperability through virtualization.
Together, these new features will let users easily transition software and data from their local computers to Bridges, and in doing so, open new, larger, and wider ranging lines of research.
One major focus of Bridges will be to allow neuroscientists to create fine maps of brain function, such as this network of 500 neural regions reconstructed from simulated magnetic resonance imaging. Some sequencing involves matching billions of short DNA sequences. (image above) [Ramsey, Joseph D., Ruben Sanchez-Romero, and Clark Glymour. (2014) Non-Gaussian methods and high-pass filters in the estimation of effective connections. Neuroimage 84:986-1006.]
BRIDGING TO THE DATA-INTENSIVE ERA
One of the greatest challenges to researchers in the 21st century is to extract knowledge from data so large and complex that its effective analysis and management requires substantial innovation. Using software developed at PSC and hardware developed by HP, Intel and NVIDIA, Bridges will provide flexible performance capabilities, combining largememory nodes that can be applied to analyze huge amounts of data at unprecedented speeds with hundreds of smaller nodes for analyses that can be partitioned. These components are supported by a high-performance data management system and shared file system linked by an interconnect architected specifically for this purpose.
|“Bridges represents a new approach to supercomputing that helps keep PSC and Carnegie Mellon University at the forefront of high-performance computing. It will help researchers tackle and master the new emphasis on data that is now driving many fields.” —Subra Suresh, President, Carnegie Mellon University||“The ease of use planned for Bridges promises to be a gamechanger. Among many other applications, we look forward to its helping biomedical scientists here at Pittand at other universities unravel and understand the vast volume of genomic data currently being generated.” —Patrick D. Gallagher, Chancellor, University of Pittsburgh|
BRIDGING TO UNIVERSITY CAMPUSES
At peak times users often swamp a university’s computational resources. A pilot project with fellow Pennsylvania institution Temple University will connect that campus with Bridges to give their researchers additional computing capacity at times of unusually high use, as well as to let Bridges offload suitable work to the Temple cluster.