TITLE:

THE DESIGN OF A PARALLEL AND DISTRIBUTED COMPUTER, A NEUROSCIENCE PERSPECTIVE (No. 61)

DATE:

Friday, October 8th, 2004

TIME:

3:30 PM

LOCATION:

GMCS 214

SPEAKER:

Faramarz Valafar, Department of Computer Science, San Diego State University

ABSTRACT:

The growing maturity of commodity platforms for high-performance computing (HPC) is expanding the impact of HPC into areas where traditionally proprietary hardware/software combinations have dominated. Commodity platforms have strongly extended their reach out of traditional uses into more complex environments requiring high degrees of scalability and reliability.

One area that has traditionally been out of reach of commodity platforms has been high-performance computing (HPC) for technical, scientific, and intensive analytical applications. Recently, however, the scientific and technical computing communities have shown greater interest in using commodity platforms in cluster architectures to deliver processor-intensive applications. Moreover, this increased processing capability is extending its reach into business-oriented application, including automotive, financial services, retail, biotech, and pharmaceuticals.

While the transition to “commodity computing” can be largely attributed to rapid decrease in cost-to-performance ration of these platforms, questions have been raised as to the scientific merits of the
distributed nature of this trend. The question at the heart of the
matter seems to be: what percentage of all applications can benefit from (be implemented on) a distributed system?

The field of computational neuroscience deals with the mechanics of how our cognitive system performs the computations that are necessary to carry out complex tasks (e.g. those performed by the human brain). Recent discoveries in computational neuroscience might be able to offer insight that could help us answer the above question.

In this talk we will discuss the over lap of the fields of computational
neuroscience and high-performance computing and look for clues in the brain to improve our high-performance distributed computing technology

HOST:

Paul Paolini

DOWNLOAD: