In work since October 2010, when Blacklight became available for NSF allocations, it has enabled advances in fields that include nanomaterials, genomics, machine learning, astrophysics, geophysics, natural language processing and climate modelling.
"As we expected it would, Blacklight has opened new doors to high-performance computation in many research communities", stated PSC scientific directors Michael Levine and Ralph Roskies, "and rapidly become a force across a wide and interesting spectrum of fields."
"Blacklight represents the leading edge in technical computing", stated SGI CEO Mark J. Barrenechea. "With this platform, the scientific community will find answers to some of the toughest questions and make new discoveries with significant impact throughout the world for years to come. We are delighted to work with the Pittsburgh Supercomputing Center to advance these efforts."
For astrophysicists Tiziana Di Matteo and Rupert Croft, Blacklight has revolutionized discovery from large-scale simulations of how the cosmos evolves. The ability to hold an entire snapshot of their MassiveBlack simulation - between three and four terabytes of data - in memory at one time was instrumental in their ability to reveal "cold gas flows" as a phenomenon that accounts for supermassive black holes in the early universe, resolving what had been a puzzle in the Cold Dark Matter model of the universe.
In a large geophysics project, a team of physicists used Blacklight to produce scientific visualizations that made it possible to see a fundamental phenomenon of space weather called magnetic reconnection, which can disrupt satellites, spacecraft and power grids on Earth. The researchers used XSEDE resources - Kraken at the National Institute for Computational Sciences, University of Tennessee, Knoxville - for very large simulations that characterize how turbulence within sheets of electrons generates structures, called "flux ropes" - that play a large role in magnetic reconnection.
"One run can generate more than 200 terabytes", stated physicist Homa Karimabadi of the University of California, San Diego. "Blacklight's shared-memory architecture is critical for analysis of these massive datasets."
In genomics, Blacklight has helped to open a potential bottleneck in processing of next-generation sequencing data. In one project, for instance, involving billions of 100-base reads from a sequencer, Blacklight's shared-memory architecture - along with consulting help from XSEDE's Extended Collaborative Support Services staff - made it possible to complete a de novo assembly in weeks, progress that had eluded James Vincent of the University of Vermont and colleagues in the Northeast Cyberinfrastructure Consortium for nearly a year in work with other systems.
With limitless quantities of text available on World Wide Web, Blacklight's shared memory provides a powerful tool for natural language processing (NLP) - sifting through billions and billions of words in various applications, including automated translators, and innovative predictive modeling. Noah Smith of Carnegie Mellon University produced four studies in diverse areas of NLP within six months of access to Blacklight.
"Blacklight has been a very useful resource for us", stated Noah Smith. "We can incorporate deeper ideas about how language works, and we can estimate these more complex models on more data."
For more information about these projects and others on Blacklight, you can visit PSC's Projects in Scientific Computing at http://www.psc.edu/science/2011
To help researchers take advantage of Blacklight, PSC provides a Memory Advantage Programme to develop applications that can effectively use Blacklight's shared-memory capabilities. These include rapid expression of algorithms - such as graph-theoretical software, for which distributed memory often presents obstacles, and interactive analysis of large data sets, which often can be loaded in their entirety into Blacklight's shared memory. For such projects, a PSC consultant can provide advice on debugging, performance-analysis and optimizations.
In computer terms, "shared memory" means a system's memory can be directly accessed from all of its processors, as opposed to distributed memory - in which each processor's memory is directly accessed only by that processor. Because all processors share a single view of data, a shared memory system is, relatively speaking, easy to program and use.