Danny Weyns - An Introduction to Self-adaptive Systems: A Contemporary Software Engineering Perspective (IEEE Press)

1. Chapter 2
2. Chapter 4
3. Chapter 5
4. Chapter 6
5. Chapter 7
6. Chapter 8
7. Chapter 9
8. Chapter 10

1. Chapter 1
2. Chapter 2
3. Chapter 3
4. Chapter 4
5. Chapter 5
6. Chapter 6
7. Chapter 7
8. Chapter 8
9. Chapter 9
10. Chapter 10
11. Chapter 11

Danny Weyns
Katholieke Universiteit Leuven, Belgium

This edition first published 2021

2021 John Wiley & Sons Ltd

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.

The right of Danny Weyns to be identified as the author of this work has been asserted in accordance with law.

Registered Offices

John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA

John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK

Editorial Office

The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK

For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.

Wiley also publishes its books in a variety of electronic formats and by printondemand. Some content that appears in standard print versions of this book may not be available in other formats.

Limit of Liability/Disclaimer of Warranty

While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

Library of Congress CataloginginPublication Data applied for,

Hardback ISBN: 9781119574941

Cover Design: Wiley

Cover Image: Takeshi.K/Getty Images

To Frankie

From the earliest days of computing, theorists recognized that one of the most striking aspects of computation is its potential ability to change itself: rather than presenting users with a fixed set of computations determined at deployment, the system could at runtime modify both what it computes and how it computes it. However, while selfmodification was perhaps interesting from a theoretical point of view, few programming systems and engineering methods embraced this capability the advantages of doing so were not obvious given the additional complexity of reasoning about system behavior and the potential for inadvertently making a really big mess of things.

Over the past decade, however, selfadaptive systems have emerged as a fundamental element of modern software systems. Virtually all enterprise systems have builtin adaptive mechanisms to handle faults, resource management, and attacks. Increasingly systems are taking over tasks otherwise performed by humans in transportation (automated driving), medicine (assisted diagnosis), environmental control (smart buildings), and many others.

In the broad field of software engineering these changes have been mirrored in a number of seismic shifts. The first has been a shift in focus from development time to runtime. For most of its history, software engineering primarily focused on getting things right before a system was deployed. When problems were encountered, systems were taken offline and fixed before redeployment. This made sense because few systems required nonstop availability and hence they could be taken down for scheduled maintenance. But today almost all public facing systems must be continuously available, requiring that systems be modifiable (either automatically or by developers) while they continue to operate.

A second shift has been the increasing level of uncertainty that accompanies modern systems. In the past, software was typically developed for a known environment using components that were largely under the control of the developers. Today systems work in much more uncertain contexts: loads can change dramatically; resources (such as network bandwidth) can vary substantially for mobile computing; faults can arise from interaction with other systems and resources outside the control of the developer; and attacks can emerge in unexpected ways.

A third shift has been an interest in automation to reduce the cost of operations. In the 1980s it was recognized that while the cost of acquiring or developing increasingly complex computing systems was steadily declining, the overall cost of ownership was rising. The reason for this was the need for system administration, which was taking up a larger and larger fraction of the IT operational budget. By automating many of the routine functions performed by administrators, systems would become more affordable. Moreover, arguably, for today's complex software systems, complete human oversight and control is simply not possible.

A fourth shift has been the commoditization of AI. Whereas for much of its existence AI had largely been relegated to special niches (e.g., robotics), the increasing availability of planners, machine learning, genetic algorithms, and gametheoretic decision systems has made it possible to harness sophisticated reasoning and learning mechanisms in support of automation.

All of these shifts have led to a set of critical challenges for software engineers. What are the fundamental unifying principles that underlie selfadaptive systems? How should one go about engineering them in a way that allows us to assure the system matches its requirements even as those requirements change after deployment? How can we provide safeguards against adaptationgoneawry? How do we engender trust in systems where human oversight has been delegated to the machine? How do we decompose the engineering effort of selfadaptive systems into manageable subtasks? How can we reuse elements of one adaptive system when constructing another?

The software engineering discipline of selfadaptive systems attempts to answer these questions. It seeks to provide the principles, practices, and tools that will allow engineers to harness the vast potential of adaptation for engineering today's systems.