Brigham Young University BYU ScholarsArchive Theses and Dissertations 2003-09-16 A Descriptive Performance Model of Small, Low Cost, Diskless Beowulf Clusters Curtis R. Nielson Brigham Young University - Provo Follow this and additional works at: https://scholarsarchive.byu.edu/etd Part of the Construction Engineering and Management Commons, and the Manufacturing Commons BYU ScholarsArchive Citation Nielson, Curtis R., "A Descriptive Performance Model of Small, Low Cost, Diskless Beowulf Clusters" (2003). Theses and Dissertations. 98. https://scholarsarchive.byu.edu/etd/98 This Thesis is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of BYU ScholarsArchive. For more information, please contact [email protected], [email protected]. A DESCRIPTIVE PERFORMANCE MODEL OF SMALL, LOW COST, DISKLESS BEOWULF CLUSTERS by Curtis R. Nielson A thesis submitted to the faculty of Brigham Young University in partial fulfillment of the requirements for the degree of Master of Science School of Technology Brigham Young University December 2003 Copyright © 2003 Curtis R. Nielson All Rights Reserved BRIGHAM YOUNG UNIVERSITY GRADUATE COMMITTEE APPROVAL of a thesis submitted by Curtis R. Nielson This thesis has been read by each member of the following graduate committee and by majority vote has been found satisfactory. Date C. Richard G. Helps, Chair Date Joseph J. Ekstrom Date Michael G. Bailey BRIGHAM YOUNG UNIVERSITY As chair of the candidate’s graduate committee, I have read the thesis of Curtis R. Nielson in its final form and have found that (1) its format, citations, and bibliographical style are consistent and acceptable and fulfill university and department style requirements; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the graduate committee and is ready for submission to the university library. Date C. Richard G. Helps Chair, Graduate Committee Accepted for the Department Thomas L. Erekson Director, School of Technology Accepted for the College Douglas M. Chabries Dean, College of Engineering and Technology ABSTRACT A DESCRIPTIVE PERFORMANCE MODEL OF SMALL, LOW COST, DISKLESS BEOWULF CLUSTERS Curtis R. Nielson School of Technology Master of Science Commodity supercomputing clusters known as Beowulf clusters, have become a low cost alternative to traditional supercomputers. Beowulf clusters combine inexpensive computers and specialized software to achieve supercomputing power. The processing nodes in a diskless Beowulf cluster do not have a local hard disk unlike the nodes in most commodity clusters. Research has provided performance information for diskless clusters built with expensive, high performance equipment. Beowulf clusters use commodity off-the-shell hardware, and little information is available about their performance. This research includes the construction of several diskless Beowulf clusters. Using the NAS Parallel Benchmarks, the performance of these clusters was measured. Through analysis of these measurements, a descriptive performance model of diskless Beowulf clusters was produced. ACKNOWLEDGEMENTS There are many individuals whose support and encouragement during this thesis I am deeply grateful for. I would to like express appreciation to my committee chair, Richard Helps, for the many hours of diligent assistance he has provided. I owe so much to my dear wife, Vicky, whose support through frustrating days and sleepless nights was always selflessly given. I would like to thank all of my family and friends who have supported me through this and so many other important steps in my life. Table of Contents 1 INTRODUCTION ..................................................................................1 1.1 Background......................................................................................................... 1 1.2 Problem Statement.............................................................................................. 3 1.3 Hypothesis........................................................................................................... 5 1.4 Justification......................................................................................................... 6 1.5 Methodology....................................................................................................... 6 1.6 Assumptions........................................................................................................ 7 1.7 Delimitations....................................................................................................... 7 2 REVIEW OF LITERATURE................................................................9 2.1 Supercomputing .................................................................................................. 9 2.1.1 Mainframe Computing....................................................................................... 9 2.1.2 Metacomputing ................................................................................................ 10 2.1.3 Clustered Computing ....................................................................................... 12 2.2 Beowulf Clusters............................................................................................... 13 2.3 Diskfull vs. Diskless Clusters ........................................................................... 16 2.3.1 Dell Diskfull vs. Diskless Study ...................................................................... 17 2.3.2 Testing Performance with the HINT Benchmark ............................................ 19 2.3.3 Testing Performance with the HPL Benchmark .............................................. 20 2.3.4 Diskless vs. Diskfull Cluster Conclusions....................................................... 22 2.4 Message Passing............................................................................................... 23 2.5 Message passing Libraries ................................................................................ 24 2.6 MOSIX.............................................................................................................. 26 2.7 Message Passing vs. MOSIX........................................................................... 27 2.8 Beowulf Cluster Software................................................................................. 29 2.9 Existing Performance Metrics........................................................................... 30 2.9.1 Speed-up .......................................................................................................... 30 2.9.2 Efficiency......................................................................................................... 31 2.9.3 Theoretical Peak Performance ......................................................................... 32 2.9.4 Application Performance ................................................................................. 32 2.9.5 Application Run Time...................................................................................... 34 2.10 Supercomputing Performance Benchmark Suites ............................................ 34 2.10.1 TPC Benchmarks ........................................................................................... 37 2.10.2 Ping–Pong Test .............................................................................................. 38 2.10.3 The Perfect Benchmarks................................................................................ 38 2.10.4 Livermore Fortran Kernels ............................................................................ 39 2.10.5 The Linpack Benchmark................................................................................ 39 2.10.6 The NAS Parallel Benchmark Suite .............................................................. 40 2.11 Review of Literature Conclusions .................................................................... 40 vii 3 RESEARCH PROCEDURES............................................................. 43 3.1 Diskless Cluster Construction........................................................................... 43 3.2 Diskless Node Boot Requirements ................................................................... 44 3.3 Diskless Node Kernel ....................................................................................... 45 3.4 Booting a Diskless Node................................................................................... 47 3.5 Dynamic Host Configuration Protocol (DHCP)............................................... 49 3.6 Network File System (NFS).............................................................................. 49 3.7 Diskless over NFS............................................................................................. 50 3.8 Diskless Node Startup Scripts........................................................................... 52 3.9 Secure Shell (SSH) ........................................................................................... 53 3.10 Local Area Multicomputer (LAM) ................................................................... 54 3.11 Diskless Beowulf Cluster Configuration Procedure......................................... 56 3.12 The NAS Parallel Benchmark Suite ................................................................. 58 3.12.1 BT Benchmark..............................................................................................