Data-Intensive Computing Initiative (DICI)
Demonstration: Scientific Discovery & Insight
The demonstration for the Scientific Discovery and Insight (SDI) Focus Area addresses the management of large volumes of experimental data that result from various technologies and the processing of these data to detect and extract features of interest. Approaches to SDI problems have historically generated solutions unique to each problem, resulting in considerable expense with minimal impact. PNNL's novel approach will enable the "toolbox" of developed capabilities to map to many domains and allow developers to more effectively and efficiently address future data-intensive problems.
Goals:
- Develop a near-real-time diagnostic capability that analyzes data from multiple data streams and supports a timely decision in the field.
- Develop a biological analysis workflow that provides analysis capabilities to the scientist.
Approach:
- FPGAs used to acquire high-speed streaming data, providing real-time data reduction and feature extraction.
- High-performance feature extraction and characterization from raw data using advanced computational algorithms.
- Application of multithreaded architectures and cell processors to enable minimal time-to-solution.
- Adaptable workflows supporting real-time diagnostic applications and scientific problem-solving environments.
The SDI demonstration is planned in three phases, shown here in green (I), gray (II), and blue (III).
Projects that support this demonstration in Phase I:
- Hybrid Algorithms for Networked System Analysis
Projects that support this demonstration in Phase II:
- MeDICI Architecture (Adaptive Software Architecture), (A Data Virtualization Architecture), (Adaptive Workflow in Data-Intensive Environments)
- Data-Intensive Machine Learning for Real-Time Decision Analysis
Projects that support this demonstration in Phase III:
- MeDICI Architecture (Adaptive Software Architecture), (Adaptive Workflow in Data-Intensive Environments)
- Data-Intensive Machine Learning for Real-Time Decision Analysis
- Multithreaded Architectures for DI Computing Applications
- Adaptive Network Traffic Analysis on the Cell Processor
- Adaptive Composite Analysis for Complex Systems
- Integrated Demonstrations of Biological Workflows
This demonstration is being developed in collaboration with PNNL's Environmental Biomarkers Initiative.