Photo credit: Loews Resort, Annapolis Maryland
|
CLSAC 2014
October 14--16, 2014 Loews Resort, Annapolis Maryland |
Large-scale data analytics is increasingly dominating government mission work, business intelligence, and scientific discovery. There has been significant change and investment in massively scalable systems over the last decade in the commercial sector (e.g. Google, Amazon, Facebook, Yahoo) to enable large-scale analytics. More specialized high performance computing has been an ongoing activity for decades within the Science, Energy, and Defense sectors. When does it make sense to employ HPC for large scale analytics in addition to large scale commodity installations? What are the recommended architectures, software stacks, file systems or schedulers? What are the right measures of performance to understand the cost/performance tradeoffs of alternative platforms?
Assembling heterogeneous information from a variety of sources provides a comprehensive view of complex systems enabling deep insights and improved decision-making capabilities. Large-scale data analytic applications process large, dynamically changing data sets by searching the data set to select key features, find patterns of related items, and print or visualize the results in real-time. Currently deployed systems comprise a mixture of loosely connected servers, storage systems, and networks. If selected features are joined to reveal relationships between different individuals, their preferences, and anomalous behaviors, then more tightly coupled, high-performance, high-bandwidth systems are required.
Organizing Committee:
Jim Ang (SNL)
John Feo (PNNL)
Dave Mountain (DoD)
Richard Murphy (Micron Technology)
Ron Oldfield (SNL)
Steve Pritchard (DoD)
Assembling heterogeneous information from a variety of sources provides a comprehensive view of complex systems enabling deep insights and improved decision-making capabilities. Large-scale data analytic applications process large, dynamically changing data sets by searching the data set to select key features, find patterns of related items, and print or visualize the results in real-time. Currently deployed systems comprise a mixture of loosely connected servers, storage systems, and networks. If selected features are joined to reveal relationships between different individuals, their preferences, and anomalous behaviors, then more tightly coupled, high-performance, high-bandwidth systems are required.
Organizing Committee:
Jim Ang (SNL)
John Feo (PNNL)
Dave Mountain (DoD)
Richard Murphy (Micron Technology)
Ron Oldfield (SNL)
Steve Pritchard (DoD)
LOGOS of SPONSORS (perhaps in the footer)
Agenda
| agenda_2014_clsac.pdf |