Computer Science ETDs

Publication Date

12-1-2011

Abstract

Many scientific fields and commercial industries need to solve computationally large problems. These problems can often be broken into smaller tasks, which can be executed in parallel, on large computer systems. In pursuit of solving ever larger problems in a more timely manner, the number of nodes in these large computer systems have grown to extremely large scales. All of the extreme-scale-software systems that run on these extreme-scale-computational systems go through what we call a bootstrapping phase. During this phase, the software system is deployed onto a set of computers and its initialization information is disseminated. In this thesis, we present a framework called the Lightweight Infrastructure-Bootstrapping Infrastructure (LIBI) to support extreme-scale-software systems during their bootstrapping phase. The contributions of this thesis are as follows: a classification system for process-launching strategies, an algorithm for creating an optimal-process-launching strategy, an implementation of LIBI and a performance evaluation of LIBI. Our performance evaluation demonstrates that we decreased the time required for software system bootstrapping by up to 50%.

Language

English

Keywords

high-performance computing, scalable systems, process launching, Lightweight Infrastructure-Bootstrapping Infrastructure

Document Type

Thesis

Degree Name

Computer Science

Level of Degree

Masters

Department Name

Department of Computer Science

First Advisor

Arnold, Dorian

First Committee Member (Chair)

Arnold, Dorian

Second Committee Member

Bridges, Patrick

Third Committee Member

Tapia, Lydia

Fourth Committee Member

Ahn, Dong

Fifth Committee Member

Lee, Greg

Project Sponsors

Lawrence Livermore National Security, LLC

Share

COinS