High-performance data intensive computing architectures are mature technologies that are well established in many commercial applications.
Most scientific super-computing however, is still done on parallelized server-based machines.
Data is moved to the CPU.
The data transfer delays make parallelized server-based approaches inefficient for data intensive problems.
Now, massively parallel field programmable gate array (FPGA)-based database appliances parallelize both data storage and processing. Computation is done at the data.
This is a new paradigm in scientific computing. Data intensive computing can provide completely new ways of solving difficult problems and solve problems that are not possible to solve in reasonable time on other architectures.
High performance data intensive computing architectures have the potential to transform many scientific computing problems. However, there is a dearth of research in this area.
DI2 focuses on meeting the unmet need for data intensive computing science
and engineering, algorithm and applications development.