High-performance cluster computing, algorithms, implementations and performance evaluation for computation-intensive applications to promote complex scientific research on turbulent flows

Thumbnail Image
Date
2001-01-01
Authors
Wang, Hao
Major Professor
Advisor
G. M. Prabhu
E. S. Takle
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Authors
Research Projects
Organizational Units
Organizational Unit
Journal Issue
Is Version Of
Versions
Series
Department
Computer Science
Abstract

Large-scale high-performance computing is a very rapidly growing field of research that plays a vital role in the advance of science, engineering, and modern industrial technology. Increasing sophistication in research has led to a need for bigger and faster computers or computer clusters, and high-performance computer systems are themselves stimulating the redevelopment of the methods of computation. Computing is fast becoming the most frequently used technique to explore new questions. We have developed high-performance computer simulation modeling software system on turbulent flows. Five papers are selected to present here from dozens of papers published in our efforts on complex software system development and knowledge discovery through computer simulations. The first paper describes the end-to-end computer simulation system development and simulation results that help understand the nature of complex shelterbelt turbulent flows. The second paper deals specifically with high-performance algorithm design and implementation in a cluster of computers. The third paper discusses the twelve design processes of parallel algorithms and software system as well as theoretical performance modeling and characterization of cluster computing. The fourth paper is about the computing framework of drag and pressure coefficients. The fifth paper is about simulated evapotranspiration and energy partition of inhomogeneous ecosystems. We discuss the end-to-end computer simulation system software development, distributed parallel computing performance modeling and system performance characterization. We design and compare several parallel implementations of our computer simulation system and show that the performance depends on algorithm design, communication channel pattern, and coding strategies that significantly impact load balancing, speedup, and computing efficiency. For a given cluster communication characteristics and a given problem complexity, there exists an optimal number of nodes. With this computer simulation system, we resolved many historically controversial issues and a lot of important problems.

Comments
Description
Keywords
Citation
Source
Subject Categories
Copyright
Mon Jan 01 00:00:00 UTC 2001