Fair Scheduling in High Performance Computing Environments
This book introduces a new scheduler to fairly and efficiently distribute system resources to many users of varying usage patterns compete for them in large shared computing environments. The Rawlsian Fair scheduler developed for this effort is show
- PDF / 5,980,078 Bytes
- 136 Pages / 439.42 x 683.15 pts Page_size
- 62 Downloads / 245 Views
ir Scheduling in High Performance Computing Environments
Fair Scheduling in High Performance Computing Environments
Art Sedighi • Milton Smith
Fair Scheduling in High Performance Computing Environments
Art Sedighi Industrial, Manufacturing & Systems Engineering Texas Tech University Lubbock, TX, USA
Milton Smith Industrial, Manufacturing & Systems Engineering Texas Tech University Lubbock, TX, USA
ISBN 978-3-030-14567-5 ISBN 978-3-030-14568-2 https://doi.org/10.1007/978-3-030-14568-2
(eBook)
Library of Congress Control Number: 2019934954 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This book introduces a new scheduler—the Rawlsian Fair scheduler—which can distribute resources fairly in situations in which users with different usage profiles are competing for resources in a large, shared computing environment. The Rawlsian Fair scheduler is demonstrated to increase performance and reduce delay in highperformance computing workloads of four classes: Class A – Similar but complementary workloads Class B – Steady vs. intermittent workloads Class C – Large vs. small workloads Class D – Large vs. noise-like workloads The Rawlsian Fair scheduler achieves short-term fairness in cases in which varying workloads and usage profiles require rapid responses. It is shown to consistently benefit workloads in C and D situations and to benefit workloads of disproportionate sizes in A and B situations. This dissertation also presents a new simulation framework—dSim—which was created to simulate the new Rawlsian Fair scheduler. A series of simulations performed using dSim demonstrate that the Rawlsian Fair scheduler makes scheduling decisions that can ensure i
Data Loading...
