Young H. Cho (young@myri.com)

Statement of Purpose/Research Plan

Updated on November 1, 1998

INTRODUCTION

From the time I was in the third grade, I have wanted to become a computer scientist. While I have been pursuing my goal for the last fifteen years, the computer industry has grown at an astoundingly rapid rate. This industrial growth brought forth thousands of new computer hardware and software implementations. Among these implementations, the successful ones are packaged as simple building blocks for complex systems. Computer companies are competing to utilize these readily available and inexpensive components to build high performance systems at low cost.

PROBLEM

Ideally, defining the most simple, yet flexible set of computer architectural rules is crucial for the future success of the computer industry. Sets of systematic rules can be applied to re-integrate existing systems and to develop new building blocks to yield superior computer design. However, finding such architecture is becoming difficult due to the increasing number of components, each with its own unique micro-architecture. Combining these parts allow even more possibilities of different architectures. Without an optimal architectural rule, future systems will not only become too complex and specialized, yet incompatible across different architectures. Then the computer industry is stuck with a problem that becomes harder to solve as time passes.

THESIS

Human beings generally accept the assumption that we are the most intelligent and adaptable organism. Effectively, I believe the human, as a system, is the most efficient and expandable system/creation. I hypothesize that a computer system modeled after brain of human or other less advanced animals would define a very efficient, if not the best, architecture. I also believe that testing this hypothesis will reveal better ideas for designing efficient building blocks. My initial research will incorporate some main biological concepts into the existing research machines built using off-the-shelf components. Therefore, the experiments will be relatively safe and cost-effective.

SPECIALIZATION

There are many aspects of computer architecture that are in need of more research and improvement. Areas that I am most interested in are Network Architecture and Parallel Computing. Compared to its biological counter-part, current computer network and processor architecture is deficient in many areas including bandwidth, connectivity, memory distribution, fault-tolerance, and power-consumption. Depending on time and resource, my research will be narrowed down further into either field.

RESEARCH PLAN

My research will begin with finding current solution that yields the highest efficiency in network and processor architecture. All such systems whould be built using readily available and cost-effective components; like various cluster systems with high-speed networks (i.e. Berkeley NOW, Princeton SHRIMP, RWCP cluster computers, and etc.) and various SMPs. Once I select a system or combination of systems, I will develop simulation models based on the findings. The system simulation will be built for easy and flexible monitoring of the performance and reconfiguration. I will research and study neural sciences, focusing on the field of connectivity and brain structure. Once I have collected sufficient information, I will incorporate biological concepts coupled with existing solutions into the simulation models. At that point, I will extract the architectural rules that yield best solutions from the simulation results.

By the time I have extracted the architectural rules, I expect to have discovered many ideas that can be modified to suit my biological network and computing models. If resources allow further research, I will build actual hardware and software system to verify my findings. The the next step will be to produce custom building blocks with requirements to house the biological architecture (i.e. network interface boards and/or parallel processors.)