ITR - ASE - [int+dmc]: Storage Based Supercomputing for Scientific Applications

Washington University, Saint Louis MO

Investigators

Abstract

ITR - ASE - [int+dmc]: Storage Based Supercomputing for Scientific Applications Abstract In many scientific computing applications the size of the data sets, already in the range of multiple terabytes, continues to grow rapidly. The time required to move this data from magnetic media, across a system bus into main memory, and then execute code to perform a computation is prohibitive. We are addressing the bottleneck by moving the computations to be implemented close to the magnetic media in fast, customizable hardware tailored to the application. We will explore the benefits associated with our proposed architecture on two specific application domains: genomics and astrophysics. In the genomics domain we will explore new approaches to discovering similarities between biological sequences. In the astrophysics domain we will examine the processing of gamma ray burst data. The results of this research will directly translate into more capability in scientific computations that are dependent upon large data sets. This will have a direct positive impact on the fields of astrophysics and genomics. The impact, however, will go well beyond these domains since the tools, techniques and architecture developed will be applicable to a wide range of scientific computing problems. We are developing a computing system that allows effective exploration of the huge and growing datasets that are central to modern scientific discovery. This new computer architecture is a central and critical instrument, analogous to the telescope and microscope, in moving forward scientific boundaries.