Internetware a new software paradigm for internet computing

The indispensable Internet poses many new challenges tosoftware engineering; these challenges are becoming particularly acute withthe emergence of computing applications that exploit the

Internetware Hong Mei · Jian Lü

A New Software Paradigm

for Internet Computing

Internetware

Hong Mei • Jian Lü

Internetware

A New Software Paradigm for Internet Computing

123

ISBN 978-981-10-2545-7 ISBN 978-981-10-2546-4 (eBook)

DOI 10.1007/978-981-10-2546-4

Library of Congress Control Number: 2016956187

© Springer Science+Business Media Singapore 2016

This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, speci fically 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 speci fic 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.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer Nature Singapore Pte Ltd.

The registered company address is: 152 Beach Road, #22-06/08 Gateway East, Singapore 189721, Singapore

Foreword I

The Internet has become both the nervous system that coordinates our societyand the brain in which all human knowledge is kept and accessed It changedfrom being a network of networks enabling access to remote machines to anetwork of content, applications, people, and (software) services, therebyweaving itself into the fabric of today’s global and interconnected physicalworld and society The indispensable Internet poses many new challenges tosoftware engineering; these challenges are becoming particularly acute withthe emergence of computing applications that exploit the Internet in new

computing, sensor networks, cyber-physical systems, data analytics, and theInternet of Things is pulling software engineering further and further from thecomfort zone of principles and techniques that have prevailed for decades.This book presents results of an ambitious plan to meet the challengesparadigmatically, i.e., to systematically re-examine the fundamental aspects

of WHAT-IS, HOW-TO-DO, HOW-TO-RUN, and HOW-WELL for ware systems situated in the open and dynamic Internet environment.Prof Hong Mei and Prof Jian Lü led the research and coined the word

soft-“Internetware” to name the novel software paradigm for Internet computing.Technically, the paradigm is intended to support the construction, operationand evolution of software systems that consist of self-contained, autonomousentities situated in distributed nodes of the Internet and coordinators con-necting these entities statically and dynamically in various kinds of inter-action styles These systems are also able to perceive the changes of open anddynamic environment, respond to changes through architectural transfor-mations, and exhibit context-aware, adaptive and trustworthy behaviors inthe Internet environment The book gives a systematic and complete cov-erage of all the related aspects of Internetware, from the programming model

to the engineering methods, to the runtime platforms, to quality assurance,and to the real-world practice and experiences It also contains many timelyand valuable new ideas, concepts, and technologies that will attract theinterest of both academia and industry, fostering further advancement ofresearch and developments in this area

In addition to its technical contributions, the successful story of netware also exemplifies how original and systematic innovations can beachieved with deep insights, strong leadership, and persistent efforts This isespecially impressive considering the fact that Internetware was conceived

Inter-v

independently in China 15 years ago when the software engineering

com-munity was predominantly representing professionals and researchers from

Western countries

This is an outstanding book on the new software engineering for the

Internet computing: it is not only a summary of research efforts by the

authors, but rather an accessible foundational reference book on software in

the contemporary age

Carlo GhezziPolitecnico di Milano

Milan, Italy

Foreword II

The convergence of various network technologies and applications, such assensor networks, cellular networks, and social networks have pushed majorInternet evolutions Internet has evolved from a content delivery and resourcesharing platform towards a ubiquitous infrastructure connecting humansociety, physical world, and massive computation resources and facilities.Software systems and applications running in such an “Internet-centric”environment are built upon highly virtualized computing environments, andhave to continuously provide high-quality services and experiences for userswho access applications from anywhere, at anytime, and by any device As aresult, software engineering development processes in all their phases andactivities, such as software development, deployment, operation, mainte-nance, and evolution, are facing major challenges arising from the openness,situation-awareness, changeability, interactivity, adaptability, and evolu-tionary nature of the Internet-based applications It is clear that addressingsuch challenges requires rethinking the principles and methods of softwareengineering

A group of distinguished Chinese researchers proposed the term netware” to refer to the new software paradigm for Internet computingenvironments They devoted intense research efforts towards addressingsignificant fundamental issues for Internetware, including the autonomy ofsoftware entities, the openness of collaboration among software entities, theevolution and adaptation driven by context changing, and the trustworthiness

“Inter-of the system behavior For almost 15 years, this group “Inter-of researchersinvestigated approaches to address these issues with the sponsorship ofNational Basic Research Program of China (“973 program”) Their missionincluded studying, establishing and using systematic, disciplined, quantifi-able approaches for software development, operation, and maintenance onInternet Internetware is now consolidated as a quite full-fledged frameworkconsisting of a new programming model, engineering approaches, runtimeplatforms, and quality assurance for software systems and applications run-ning on the Internet

Research on Internetware has resulted in numerous results recognized bypremier conferences, top journals, real-world applications, and industrialsolutions and products This book, titled Internetware: A New SoftwareParadigm for Internet Computing, summarizes over 15 years of researcharound software engineering for Internet-based applications and the valuable

vii

results of this research The book presents a core set of principles and

methodologies for Internetware that will help engineers and designers in

enhancing the effectiveness and productivity of the development processes

for Internetware applications The book also provides comprehensive

guidelines and many representative real-world case studies that will be

invaluable for software engineers in the development of Internetware

applications

This book also provides an important analysis of current research trends in

modern software engineering in the Internet era Many ideas and results

discussed in the book represent important research directions that can be

further extended and developed into more solid and theoretical foundations

of software engineering for Internet computing

I believe that this book is an important resource in thefield of software

engineering for everyone interested in novel trends, research, and

method-ologies in today software engineering

Enjoy the book!!

Elisa BertinoPurdue UniversityWest Lafayette, USA

Foreword III

We are undergoing a tremendous change towards a new Internet-based andsoftware-defined cyberspace, where our physical world and social commu-nities can be virtualized by software technologies, delivered as softwareapplications and services, and operated over an infrastructure that isInternet-based and globally networked Cloud-based, mobile-oriented, andsensor-laden software applications and services establish the ubiquitousfabrics among every object in such a cyberspace, and have to adapt toever-changing and on-the-fly requirements in response to instant feedbackfrom this Internet-scale computing environment that is becoming more andmore complex As such, software engineering tasks are increasingly complexand highly challenging in view of unstable, noisy, and sometimes harshenvironments However, current software engineering approaches and tech-niques largely lack the capabilities to continuously explore the collaborative,situation-aware, adaptational, and evolutionary needs of software applica-tions and services Software engineering research needs substantial evolu-tionary or even revolutionary improvements

The novel Internetware research in China is such an inspiring example.Led by Prof Hong Mei and Jian Lü, as well as numerous other researchers inChina, this area has made great efforts towards revolutionizing the emerging

“Internet-based” and “software-defined” cyberspace in the past 15 years TheInternetware is proposed as a new software paradigm covering new pro-gramming abstractions, engineering approaches, runtime infrastructures, andquality assurance to make software applications and services collaborative,adaptive, evolvable, situational, emergent, and trustworthy Internetwareresearch has brought low-hanging fruits in premier academic conferences andjournals Meanwhile, some of the results have been successfully applied inreal-world applications In my opinion, Internetware research not only leadsthe pioneering software research and development in China, but also makessignificant impacts on the software community worldwide

This book provides to readers the authors’ unique perspectives and effortsfrom the Internetware perspective, including the programming models,engineering approaches, middleware platforms and operating systems,measurements and quality-of-experience assurance, etc This book alsocontains timely and valuable new ideas, concepts, and technologies for manyapplication contexts such as cloud computing, mobile computing, socialcomputing, cyber-physical systems, and so on

ix

It is my great pleasure to introduce this must-read book to faculty and

students, researchers and practitioners, decision makers in charge of

gov-ernmental research funding and other research & development executives

who are interested in grasping and helping push the envelope of

Internet-based computing technologies

Carl ChangIowa State University

Ames, USA

To cope with new characteristics of software for the Internet computingenvironment, existing software paradigms need to be evolved to be a newone Such a new software paradigm should be able to systematically supportthe development, deployment, operation, and quality assurance of software

on the Internet Around 2000, we coined a word Internetware from the twowords“Internet” and “Software”, to denote our visions of a new paradigm forsoftware systems that are constructed, executed, used, and evolved in theopen and dynamic Internet environment Elaborately, these systems oftenneed to be

Autonomous Software entities are usually distributed, relatively contained, and independent Software entities perform according to the com-position or deployment strategies defined by their providers, and continuouslysatisfy the providers’ requirements A software entity can adapt itself whennecessary, by sensing and collecting information on environment changes.Cooperative.A set of software entities can collaborate with each other forthe purpose of business or management Often the collaborations are ratherdynamic than static, to adapt to the user requirements and environments in anon-demand way The collaboration mechanisms between software entitiescan be of various types, and can be changed if necessary

self-Situational.Software applications can be aware of the runtime contextsand scenarios, including the underlying devices, operating platforms, net-working conditions, application contexts, or the changes of other dependentapplications, etc Hence, both software entities (included in a softwareapplication) and their operating platforms might be capable of exposing theirruntime states and behaviors in some way

Evolvable.The structures and behaviors of software applications mightdynamically change Software applications usually consist of autonomousentities over the Internet, and provide online and continuous (e.g., 24 h * 7days) services for a large number of users Hence, software applicationscannot be shut down during evolution Software applications have to performonline evolution to accommodate new user requirements and environments.Possible evolutions can include addition/removal of software entities, chan-ges of functionalities on-the-fly and just-in-time, changes of interaction stylesbetween entities, change of topologies among entities, etc

xi

Emergent Software applications can exhibit undesigned behaviors or

undesired effects on runtime instances or interactions Such nature might

iteratively result in more and more changes of software application structures

and behaviors to accommodate such emergences

tradeoffs among various quality attributes As software applications serve a

number of online users, the trustworthiness of the software applications

should cover a wide spectrum, including reliability, security, performance,

user experience, etc Quality assurance can be relevant to various aspects,

including autonomous entities, interaction styles, network environments,

usage patterns, malicious attacks, software evolution, etc

Since then, substantial efforts have been made on Internetware research

and practices Especially, two national projects have been carried out

con-secutively: the project Research on Theory and Methodology of Agent-based

Middleware on Internet Platform (2002–2008) and the project Research on

researchers from Chinese universities and institutes have participated in the

projects, including researchers from Peking University, Nanjing University,

Tsinghua University, Institute of Software of the Chinese Academy Sciences,

the Academy of Mathematics and Systems Science of Chinese Academy

Sciences, East China Normal University, and IBM China Research

Labora-tory Since 2009, the Asia-Pacific Symposium on Internetware (http://sei

held annually, attracting authors and attendees from China, USA, Europe,

Australia, Japan, and Korea

Internetware research gets funding support from various China’s national

research and development programs in the past 15 years The preceding two

projects arefirst sponsored by the Chinese National Basic Research Program

(known as 973), which is one of the five major national programs in the

national R&D program of China After the first five years, some research

topics identified for Internetware, are sponsored by the National Natural

Science Foundation of China and the National High Technology Research

and Development Program (known as 863) Some prototypes of Internetware

operating platforms and development tools have been successfully

trans-ferred to commercial products and solutions under the support of the National

Science and Technology Major Projects In 2014, the Research on

Inter-netware: Theory, Approaches, and Technologies was awarded by the Top 10

Grand Progress of Science and Technology awarded by the Ministry of

Education of the PRC In a word, the Internetware research community has

been established and keeps increasing in the past a few years

Book Overview

This book is organized to summarize and share the state-of-the-art efforts of

Internetware research More specifically, this book presents the efforts that

address the challenges of software technologies in Internet computing

environment, including the fundamental aspects of programming model,engineering approach, runtime operational platform, and quality measure-ments and assurance In addition, this book also includes a number ofreal-world applications, experiences, and practices of Internetware The bookconsists offive parts.

• Part 1 gives an overview of a technical framework for Internetware Ittakes a software architecture-centric viewpoint and organizes the frame-work as three connected aspects: system modeling, middleware support,and engineering method

• Part 2 presents a software model featuring environment driven adaptationfor Internetware applications Enabling techniques for the model, espe-cially the modeling, management and utilization of environment infor-mation are discussed An architecture-centric framework has also beenbuilt to prove the concepts and evaluate the techniques

• Part 3 focuses on the runtime support for Internetware application tems The support covers various aspects of the execution and adaptation

sys-of ssys-oftware situated in the open and dynamic Internet environment

• Part 4 introduces the Internetware engineering approach It essentiallyfollows the core and underlying principle of software architecturethroughout a whole-life cycle This software architecture serves as ablueprint, and suggests or controls each phase in lifecycle for developingInternetware applications

• Part 5 describes how the Internetware paradigm is applied to real-worldcloud and client applications At the client side, applications on smart-phones connect to the physical world and Internet by their built-in sensorsand networking chips, respectively At the cloud side, numerous com-puting platforms like virtual machines and middleware infrastructures aremanaged to run user tasks in a cost-effective way We discuss how theInternetware paradigm is realized on the two sides, as well as how the twosides are connected and how applications can be diagnosed for energy

efficiency

Organizations of the Book

The book stems from a set of high-quality research papers that have beenpublished on premier computer science conferences or journals Indeed, therehave been numerous efforts contributed to Internetware research in the past

15 years Due to space limit, we have to select only some representativeefforts that cover some typical Internet-based applications such as cloud/gridcomputing, services computing, mobile computing, Web, Internet of Things,and so on These efforts are organized by the research framework of Inter-netware paradigm When having acquired the copyrights of original pub-lishers, the authors further make great efforts to revise and expand theoriginal contents to better fit the goal of this book

As mentioned previously, the Internetware research is a long-term and

open study that involves hundreds of researchers with different background

Meanwhile, the understanding, synthesis, and scope of Internetware

para-digm keep improving during our 15-year course of study The selected efforts

in this book were actually made at different time and from different

per-spectives As a result, this book cannot comprehensively enforce the

con-sistent understanding and representation of some aspects in Internetware

Indeed, it can reflect the history that we made substantive headway towards

the essence of Internetware paradigm

Audiences and Readers of the Book

This book can be a reference book to researchers and practitioners of

soft-ware engineering, operating/network systems, programming languages, and

so on It can be also used as a textbook for graduate students and

junior/senior undergraduate students who major in computer science

extension of their previous work and provided reviews of the draft script: Fuqing Yang, Gang Huang, Wei Zhang, Xuanzhe Liu, Yingfei Xiong,Haiyan Zhao, Yanchun Sun, Wenpin Jiao, Ying Zhang, Ling Lan, Junguo Li,Hui Song, Franck Chauvel, Xiaodong Zhang, from Peking University;Xiaoxing Ma, Xianping Tao, Feng Xu, Sanglu Lu, Chun Cao, Chang Xu,Wenzhong Li, Yu Huang, Ping Yu, Xiaoming Fu, Hao Hu, Tianxiao Gu,Linghao Zhang, from Nanjing University; Jun Wei, Guoquan Wu, WenboZhang, Chunyang Yue, Hua Zhong, Hong He from the Institute of Software,Chinese Academy of Sciences; Jifeng He, Jing Liu, Xiaoshan Li, ZhimingLiu, from East China Normal University; Puewei Wang, Guangjun Cai fromthe Academy of Mathematics and System Science, Chinese Academy ofScience; Lin Liu, Yang Chen, Xiaojun Ye, Yingbo Liu, from TsinghuaUniversity.

manu-Our colleagues, Tao Xie at University of Illinois Urbana-Champaign,Zhenjiang Hu at National Institute of Informatics (Japan), Shi-Chi Cheung atthe Hong Kong University of Science and Technology, were exceptionallysupportive throughout the writing of the text

Meihua Yu helped establishing the online co-editing system to improvethe work efficiency

In particular, Gang Huang (Peking University), Xiaoxing Ma (NanjingUniversity), and Xuanzhe Liu (Peking University) made great efforts to helporganize the compilation of whole book and identify the typos in early drafts

xv

Due to the limit of time, pace, our knowledge and experiences, some

possible problematic issues are inevitable in this book It would be highly

appreciated that our audiences can help point out our mistakes and improve

this book Comments and suggestions can be sent to meih@pku.edu.cn or

lj@nju.edu.cn We would like to hear from you

Thank you

1 Internetware: A Shift of Software Paradigm 3

1.1 Introduction 3

1.2 The Internetware Paradigm 4

1.2.1 The Need of a Paradigm Shift 4

1.2.2 Methodological Reflections 5

1.2.3 Architecting Principles for Internetware 7

1.2.4 Our Roadmap and Explorations 8

1.3 Open Coordination Model 8

1.3.1 Coordination Model for Internetware 8

1.3.2 A Software Architecture Based Approach 10

1.4 Environment-Driven Model 11

1.4.1 Structure and Dynamics 11

1.4.2 Framework and Techniques 13

1.5 Dependability Assurance Framework with Trust-Management 14

1.5.1 A Trust-Based Approach 14

1.5.2 A Dependability Assurance Framework with Trust Management 15

1.6 Conclusion 15

References 16

2 Technical Framework for Internetware: An Architecture Centric Approach 19

2.1 Introduction 19

2.2 Software Model of Internetware 21

2.2.1 Basic Component Model 22

2.2.2 Open Collaboration Model 22

2.2.3 Context Driven Model 23

2.2.4 Intelligent Trustworthy Model 23

2.3 Middleware for Internetware 23

2.3.1 The Implementation Model of Basic Internetware Entities 24

2.3.2 The Implementation Model of On-Demand Collaborations 24

2.3.3 The Autonomic Management of Middleware 25

2.3.4 Componentization of Middleware 25

xvii

2.4 Engineering Approach for Internetware 26

2.4.1 Software Architecture in the Whole Lifecycle 26

2.4.2 Development of Internetware Basic Entities and Their Collaborations 28

2.4.3 Domain Modeling for Internetware 28

2.5 Conclusion 29

References 29

Part II Internetware Software Model 3 On Environment-Driven Software Model for Internetware 33

3.1 Introduction 33

3.2 An Approach to the Environment-Driven Model 35

3.2.1 Overall Interaction Patterns of Environment-Driven Applications 36

3.2.2 A Structuring Model for Environment-Driven Applications 37

3.2.3 Model and System Characteristics 40

3.3 Environment Modeling and Enabling Techniques 42

3.3.1 Context Processing Framework 43

3.3.2 Ontology-Based Modeling of Dynamic Context 44

3.3.3 Logic Distance-Based Context Retrieval 46

3.3.4 Blurring Degree-Based Privacy Protection 48

3.3.5 Negotiation-Based Context Information Selection 50

3.4 Goal-Driven Adaptation and Rearon Ontologies 51

3.4.1 Rearon Ontologies 52

3.4.2 Overall Interaction Patterns of Environment-Driven Applications 55

3.5 Prototypes and Experiments 56

3.5.1 Artemis-ARC 57

3.5.2 Artemis-MAC 58

3.5.3 Artemis-FollowMeLite 60

3.5.4 Artemis-FollowMeAgent 61

3.5.5 Environment-Driven Applications 62

3.6 Related Work 64

3.7 Conclusion 66

References 67

4 On Self-adaptation Model for Internetware 71

4.1 Introduction 71

4.2 Approach Overview 73

4.3 Key Techniques 74

4.3.1 Basic SA Model 74

4.3.2 SA Reflection 76

4.3.3 SA Dynamism 76

4.3.4 SA Reasoning 77

4.3.5 SA Trustworthiness 78

4.3.6 A Tool for Modeling SA in the Whole Lifecycle 80

4.4 Case Study 81

4.4.1 Pattern Driven Self-adaptation 81

4.4.2 Style Driven Self-adaptation 85

4.4.3 Performance Impact 88

4.5 Related Work 89

4.6 Conclusion 90

References 90

5 On Requirements Model Driven Adaption and Evolution of Internetware 93

5.1 Introduction 94

5.2 Matching and Composing Internetware Components 95

5.2.1 A Component Meta Model 95

5.2.2 Component Lifecycle 97

5.2.3 Component Composition Rules 98

5.3 Components Adaptation 99

5.3.1 Adaptation Strategies 99

5.3.2 Adaptation Process 100

5.4 Internetware Testbed and Case Studies 102

5.4.1 Internetware Testbed 102

5.4.2 Model-Driven Development and Evaluation of Internetware Applications 103

5.4.3 Enterprise Applications Example 105

5.4.4 Experiment Scenarios on the Collaborations and Evolutions of Internetware Components 105

5.5 Related Work 107

5.6 Conclusion and Future Work 108

References 109

Part III Internetware Operating Platform 6 Runtime Recovery and Manipulation of Software Architecture of Component-Based Systems 115

6.1 Introduction 116

6.2 Approach Overview 118

6.2.1 Architectures in ABC 118

6.2.2 Conceptual Framework 119

6.3 Characteristics of Software Architecture at Runtime 119

6.3.1 Software Architecture at Design Time 120

6.3.2 Software Architecture at Runtime 121

6.3.3 Description for Software Architecture Recovered at Runtime 123

6.4 Recovery of Software Architecture at Runtime 124

6.4.1 Overview of PKUAS 124

6.4.2 Recovery of Basic Elements 126

6.4.3 Construction of Architecture Views 129

6.4.4 Enrichment of Semantics 130

6.5 Manipulation of Recovered Software Architecture

at Runtime 131

6.5.1 Manipulation via Reflection 131

6.5.2 Programming Model 132

6.5.3 Graphical Tool 132

6.6 Performance Evaluation 134

6.7 Related Work 135

6.8 Conclusion and Future Work 136

References 137

7 Supporting Runtime Software Architecture: A Bidirectional-Transformation-Based Approach 139

7.1 Introduction 140

7.2 Runtime Software Architecture 141

7.2.1 An Illustrative Example 141

7.2.2 A Formal Description of Runtime Software Architecture 142

7.3 Maintaining Causal Connections by Architecture-System Synchronization 144

7.3.1 The Four Properties 144

7.3.2 The Challenges 144

7.4 Architecture-System Synchronization Based on Bi-transformation 145

7.4.1 Enabling Techniques 145

7.4.2 The Synchronization Algorithm 146

7.4.3 Assumptions 149

7.4.4 Discussion About the Algorithm and the Properties 149

7.5 Generating Synchronizers for Legacy Systems 150

7.5.1 Implementing the Generic Synchronization Engine 151

7.5.2 Generating Specific XMI Parsers and System Adapters 151

7.6 Case Studies 152

7.6.1 C2-JOnAS 152

7.6.2 Client/Server-JOnAS 154

7.6.3 Other Case Studies 155

7.6.4 Summary and Discussion 155

7.7 Related Work 158

7.8 Conclusion 159

References 159

8 Low-Disruptive Dynamic Updating of Internetware Applications 163

8.1 Introduction 163

8.2 Dynamic Updating: Eager Versus Lazy 165

8.3 Javelus: Overview 166

8.4 Dynamic Patch Generation 167

8.4.1 Identifying Changed Classes 167

8.4.2 Default Transformation and Custom Transformers 168

8.5 Updating Code 170

8.5.1 Reaching a DSU Safe Point 170

8.5.2 Updating Classes 170

8.6 Updating Objects 171

8.6.1 Affected Types 171

8.6.2 Optimizing Object Validity Checks 173

8.6.3 Mixed Object Model 174

8.6.4 Transforming Objects 175

8.6.5 Continuous Updating 176

8.7 Evaluation 176

8.7.1 Experiments with Micro Benchmarks 177

8.7.2 Experiments with Tomcat and H2 179

8.8 Related Work 179

8.8.1 Dynamic Updating Systems for Java 179

8.8.2 Dynamic Updating Systems for Other Programming Languages 181

8.8.3 Dynamic Updating for Component-Based Systems 181

8.9 Conclusion 181

References 182

9 Specification and Monitoring of Data-Centric Temporal Properties for Service-Based Internetware Systems 185

9.1 Introduction 186

9.2 Motivation Example 187

9.2.1 Car Rental Application 187

9.2.2 Supply Chain Application 190

9.3 A Language for Specifying Data-Centric Properties 191

9.3.1 The Syntax of Par-BCL 191

9.3.2 Par-BCL for Data-Centric Properties 194

9.4 Monitoring Model of Par-BCL 194

9.4.1 Filtering 194

9.4.2 Parameter Binding 196

9.4.3 Verification 197

9.5 Efficient Monitoring Process 198

9.5.1 Parameter State Machine 198

9.5.2 Monitor Instance 199

9.5.3 Monitoring Algorithm 199

9.5.4 Management of Monitor Instances 200

9.5.5 Architecture and Implementation 201

9.6 Experiments 202

9.6.1 Violation Detection 202

9.6.2 Performance Evaluation 202

9.6.3 Overhead Evaluation 204

9.7 Discussion 205

9.8 Related Work 205

9.8.1 Engineering Self-adaptive System 205

9.8.2 Runtime Monitoring of Web Services 206

9.8.3 Parametric Properties Monitoring

for OO Systems 2079.9 Conclusion 207

References 208

10.1 Introduction 212

10.2 Property Detection for Pervasive Context 213

and Logical Time 21310.2.2 Consistent Global State (CGS)

and the Lattice Structure Among CGSs 21310.2.3 Specification of Contextual Properties 214

10.3 Concurrent Activity Detection in Asynchronous

Pervasive Computing Environments 215

10.3.1 Def ðφÞ and Concurrent Contextual

Activities 21510.3.2 Design of the CADA Algorithm 218

10.3.3 Discussions 219

10.4 Performance Analysis 220

10.4.1 The Causes of Faults 221

10.4.2 Concurrency Among Sensed Activities 221

10.4.3 Sensed and Physical Activities 223

10.4.4 Discussions 224

10.5 Experimental Evaluation 224

10.5.1 Implementation 224

10.5.2 Experiment Setup 225

10.5.3 Effects of Tuning the Update Interval 225

10.5.4 Effects of Tuning the Message Delay 226

10.5.5 Effects of Tuning the Duration

of Contextual Activities 227

of Non-checker Processes 22710.5.7 Lessons Learned 227

10.6 Related Work 228

10.7 Conclusion 228

References 229

for Internetware 233

11.1 Introduction 233

11.2 Overview of ABC Methodology 236

for Internetware 239

11.4 Architecture Modeling of Self-adaptive Internetware 24411.5 Reflective Middleware as Internetware

Operating Platform 24811.5.1 SA-Based Reflection 25011.5.2 Autonomous Components 25211.6 Conclusion 255References 256

Based on Environment Ontology 25712.1 Introduction 25712.2 Environment Ontology 25912.2.1 Upper Environment Ontology 25912.2.2 Domain Environment Ontology 261

Environment Ontology 262

Specification 26612.3.1 Elements of EnOnCaS 26712.3.2 An Example Capability Profile 26912.3.3 Capability Specification Generation 27012.3.4 Internetware Entity Discovery 27112.4 Related Work 27512.5 Conclusion 277References 277

and High-Level Software Design 27913.1 Introduction 27913.2 The Preliminaries 28013.2.1 Definition of Features 28113.2.2 Basic Attributes of Features 28113.2.3 Operationalization of Features 28213.2.4 Responsibility Assignment 28213.2.5 Resource Containers 28313.3 Feature Dependencies 28413.3.1 Refinement 28413.3.2 Constraint 28513.3.3 Influence 28613.3.4 Interaction 28713.4 Connections Between Dependencies 28813.4.1 Refinement-Imposed Constraints 28913.4.2 Constraint and Interaction 28913.4.3 Influence and Interaction 28913.5 Verification of Constraints and Customization 29113.5.1 Feature-Oriented Customization-Based

Requirement Reuse 29113.5.2 Customization Through a Series

of Binding-Times 291

13.5.3 Three Verification Criteria 291

13.5.4 The Effectiveness of the Three Verification

Criteria 29213.6 Design of High-Level Software Architecture 293

13.6.1 An Overview 293

13.6.2 Resource Container Identification 293

13.6.3 Component Seed Creation 294

14.7.3 Limitation and Future Work 330

References 331

Computation Offloading 337

15.1 Introduction 338

15.2 Design Pattern for Computation Offloading 339

15.2.1 The Source Structure and the Target

Structure 33915.2.2 Refactoring Steps Overview 340

15.2.3 An Illustrative Example 342

15.3 Implementation of DPartner 34215.3.1 Detect Movable Classes 34215.3.2 Generate Proxies 34315.3.3 Transform App Classes 34315.3.4 Cluster App Classes 345

to Be Offloaded 34715.3.6 Offload Computations at Runtime 34815.4 Evaluation 35015.4.1 Experiment Setup 35015.4.2 Performance of Refactoring 35115.4.3 Comparison of App Performance 351

15.4.5 The Effect of On-Demand Offloading 35415.4.6 Experiments on 3G Network 35415.5 Related Work 35515.6 Conclusion and Future Work 357References 357

of Platforms in Cloud 35916.1 Introduction 36016.2 Motivating Example 36116.3 An Architecture-Based Model for Platform

Management in Cloud 36316.3.1 Approach Overview 36316.3.2 The Architecture-Based Meta-Model 36316.3.3 Runtime Changes 365

Architecture-Based Model 36516.5 Evaluation 36616.5.1 Anti-pattern Detection 36616.5.2 VM States Checking 36816.6 Related Work 36816.7 Conclusion and Future Work 370References 371

17.1 Introduction 37417.2 System Model 37517.3 Update Propagation Strategies 37617.3.1 Random Spread 37717.3.2 Youngest Age 37717.3.3 Golden Age 37717.4 Performance Analysis 37817.4.1 Notations and Assumptions 37817.4.2 Number of Updated Nodes in the System 37817.4.3 Principles for Choosing the Golden Age 382

17.5 Numerical Results 383

17.5.1 Simulation Setup 383

17.5.2 The Number of Updated Nodes 383

17.5.3 Impact of Source Density 384

17.5.4 Impact of Buffer Size 384

17.5.5 Impact of Update Interval 385

Parameter Choosing Rules 38517.6 Related Work 385

17.7 Conclusion 386

References 387

Inefficiency for Smartphone Applications 389

18.1 Introduction 390

18.2 Background 392

18.3 Empirical Study 393

18.3.1 Problem Magnitude 394

18.3.2 Diagnosis and Bug-Fixing Efforts 394

18.3.3 Common Patterns of Energy Bugs 396

18.3.4 Threats to Validity 402

18.4 Energy Efficiency Diagnosis 402

18.4.1 Approach Overview 402

18.4.2 Application Execution and State Exploration 403

Lock Deactivation 40618.4.4 Sensory Data Utilization Analysis 407

18.5 Experimental Evaluation 413

18.5.1 Experimental Setup 414

18.5.2 Effectiveness and Efficiency 416

18.5.4 Impact of Event Sequence Length Limit 424

Detection Work 42718.5.6 Energy Saving Case Study 428

18.5.7 Discussions 429

18.6 Related Work 432

18.6.1 Energy Efficiency Analysis 432

18.6.2 Energy Consumption Estimation 433

18.6.3 Resource Leak Detection 434

18.6.4 Information Flow Tracking 435

18.7 Concluding Remarks 435

References 435

19 Conclusion and Future Outlook 441

xxvii

OOSE Object-Oriented Software Engineering

Part I Introduction to Internetware

Due to the open, dynamic, and ever-changing

nature of the Internet computing environment,

software on the Internet differs from traditional

software in terms of forms, structures, and

behaviors Consequently, software applications

(including software entities and their interactions)

for Internet computing often need to be

autono-mous, cooperative, situational, evolvable,

emer-gent, and trustworthy This puts forward many

technical challenges for software engineers

We believe that, in addition to specific research

for each concrete technical issue, a thorough

re-examination of the fundamental software

paradigm is needed A software paradigm usually

concerns with four main aspects: WHAT-IS,

HOW-WELL WHAT-IS refers to what is to be

constructed and executed (software or program

model) HOW-TO-DO refers to how to develop

the required software applications including their

constituent entities (development techniques, e.g.,

programming languages, engineering approaches,

and supporting tools) HOW-TO-RUN refers to

how to run the software applications including

their constituent entities (runtime system

sup-ports, such as the operating systems or

middle-ware platforms) HOW-WELL refers to how well

the constructed and executed software

applica-tions can perform (promised software qualities,

e.g., correctness, performance, reliability, user

experiences, and their assurance mechanisms

during software construction and operation)

One main objective of a software paradigm is tobetter leverage the underlying runtime environ-ments such as hardware capabilities while offering

a computing model that is sufficiently expressiveand natural to characterize the application domain

A software paradigm evolves with the underlyingruntime environments and the target applicationdomains For a software paradigm, WHAT-IS andHOW-TO-DO/HOW-TO-RUN are typically thefirst set of concerns to address and thenHOW-WELL becomes a concern when the soft-ware paradigm is broadly applied in practice.Every shift of software paradigms (e.g., the shiftfrom structured to object-oriented and then the

component-based/service-oriented) typically brings significant challenges

as well as tremendous opportunities to softwaretechnologies After more than a decade of devel-opment and growth, the expanded runtime envi-ronments and application domains in the era ofInternet call for a new shift on software paradigms

to cope with new requirements on software forInternet computing

In particular, a new software paradigm forInternet computing needs to satisfy the followingrequirements (organized along the four aspects;some of which are general to any software

software for Internet computing):

 Software model (WHAT-IS) The softwaremodel should specify the forms, structures,and behaviors of software entities as well as

their collaborations These specifications will

determine the principles and features of the

corresponding software technologies

(pro-gramming languages, development

software entities can be built upon current

popular technologies, such as object-oriented

technologies and service computing

technolo-gies But new capabilities should be provided

to enable on-demand collaborations among

situation-aware supports

 Software operating platform (HOW-TO-RUN)

The operating platform should provide a

run-time space to operate software entities and their

collaborations To ease the migration of legacy

software to software for Internet computing, the

operating platform should be able to

conve-niently equip legacy software with new features

added to satisfy the requirements on software

for Internet computing In addition, the

oper-ating platform should manage software

appli-cations and the platform itself in an intelligent

and automatic manner

 Engineering approach (HOW-TO-DO) The

engineering approach should systematically

control the whole lifecycle of developing

software for Internet computing, including

requirements specification, design,

implemen-tation, deployment, and maintenance

 Quality assurance (HOW-WELL) Softwareapplications on the Internet usually serve alarge number of users in an online andsimultaneous style Both quantitative andqualitative measurement methods should bedeveloped for various quality attributes such

as performance, reliability, and usability, andfor the comprehensive tradeoffs among theseattributes Assuring high quality requiresquality assurance methods realized by engi-neering approaches at the development time(e.g., testing, verification and validation), andruntime mechanisms (e.g., online evolutionand autonomic system management)

This part gives a brief introduction on themotivations and visions of the Internetwareparadigm After looking back on the driven forcesbehind pervious paradigm shifts in the history ofsoftware engineering, Chap.1discusses the need

of a new paradigm in face of the challenges ofInternet computing It also proposes a roadmaptowards this new paradigm Chapter 1 gives anoverview of a technical framework for Internet-ware It takes a software architecture-centricviewpoint and organizes the framework as threeconnected aspects: system modeling, middlewaresupport, and engineering method

integra-of Internetware applications.

Keywords

Internetware· Software paradigm · Methodological perspective

1.1 Introduction

The Internet has become a new computing

environment, in which software systems of

new styles are developed, deployed, and

exe-cuted Tremendous amount and vast diversity

of resources and services are emerging in this

environment Unprecedented potentials are seen

through emerging approaches to the coordination

and integration of these resources and services in

Parts of this chapter were reprinted from Jian Lü,

Xiaoxing Ma, Yu Huang, Chun Cao and Feng Xu.

Internetware: a shift of software paradigm, In

Proceed-ings of the First Asia-Pacific Symposium on

Internet-ware © 2009 ACM, Inc http://doi.acm.org/10.1145/

1640206.1640213.

a global scale Among these approaches there areGrid Computing [1], Service Computing [2,3],Ubiquitous Computing [4], and Cloud Computing[5], to name a few Contrasting to its traditionalcounterparts such as stand-alone computers,LANs, and enterprise intranets, the Internetcomputing environment is much more open,dynamic, and autonomous As a consequence,application systems in such an environmentcannot be constructed in a tightly-coupled andonce-for-ever manner as traditional ones often do.These new systems must respect the autonomy

of the services they use, be capable of probingthe changes of the environment and user pref-erences, and then adapt themselves accordingly

to keep their service satisfactory with sufficientfunctionality, performance, and dependability

© Springer Science+Business Media Singapore 2016

H Mei and J Lü, Internetware, DOI 10.1007/978-981-10-2546-4_1

3

4 1 Internetware: A Shift of Software …

The challenges in developing such kind

of applications not only lie in specific

tech-nical difficulties, such as inter-organization

interoperation, runtime adaptation, decentralized

authentication, etc., but also come from the lack

of suitable methods that can help developers to

understand and manage the complexity of these

applications We believe that such challenges

necessitate a major shift of software paradigm,

because there is a mismatch between the

assump-tions of conventional software paradigms and

the features of the open Internet environment

For example, conventionally, the decomposition

of a system into computing elements could be

totally under the command of its developers, but

now the autonomy of resources and services in

Internet must be respected; instead of assuming

that the working environment would be the same

as specified at development time, the system

needs to deal with constant changes in the

environment at runtime; moreover, rather than

developed within a monolithic organization, the

system is often based on independent services

with different interests

Reflections on previous shifts of software

paradigm indicate that the balance of the forces

from the problem, mental, and computing spaces

is crucial to the success of a software paradigm

New applications in new open environments

demand a new paradigm to rebalance the forces

In this chapter we discuss our visions and

explorations of the Internetware, a new software

paradigm envisioned for application systems

situated in the open Internet environment This

new paradigm emphasizes the autonomy of the

constituent entities of a system, the cooperation

and coordination among the entities, as well as

the situation-awareness, evolvability, emergence,

and trustworthiness of the system [6 9] Toward

this new paradigm, a three-staged roadmap

is proposed and adopted in our recent work

First, extending the mainstream Object-Oriented

model, we propose a new structuring model for

the coordination of autonomous services to satisfy

particular demands in a flexible but disciplined

way Second, we give an environment-driven

model to support the dynamic adaptation of

systems to the changes in environment Third,

we present a trust management-based modelfor evaluating and ensuring the dependability ofsystems built upon autonomous resources in theopen Internet environment A set of enabling tech-niques for these models are also briefly discussed.The rest of this chapter is organized as follows

In Sect.1.2we discuss the rationale and ing principles of the Internetware paradigm Sec-tions1.3,1.4, and1.5are dedicated to the opencoordination model, environment-driven model,and dependability assurance model, respectively.Finally Sect.1.6summaries the chapter and brieflydiscusses directions for future work

architect-1.2 The Internetware Paradigm

1.2.1 The Need of a Paradigm Shift

A paradigm is an accepted model or pattern digms gain their status because they are more suc- cessful than their competitors in solving a few prob- lems that the group of practitioners has come to recognize as acute To be more successful is not, however, to be either completely successful with a single problem or notably successful with any large number [10].

Para-Since the publishing of Kuhn’s seminal book

The Structure of Scientific Revolutions, the

importance of paradigms and paradigm shifts inscientific methodology is widely recognized Inhis Turing Award Lecture, Floyd convinced usthat paradigms of programming take an importantrole in helping the developers understand andmanage the complexity in solving problems [11].High level paradigms of programming-in-the-

large are also called software paradigms Various

paradigms exist in software communities Amongthem there are structured [12], object-oriented[13], component-based [14], and agent-orientedparadigms [15], to name a few In his book [16],Kaisler discussed a set of widely used softwareparadigms and their associated techniques Asoftware paradigm provides a model of howsoftware systems are developed, sometimes inthe form of a metaphor However it should beemphasized that behind the model there are acollection of assumptions, concepts, theories, andpractices acting as the rationale

1.2 The Internetware Paradigm 5

With the transformation to the open, dynamic,

and autonomous Internet environment, the

following assumptions of conventional software

paradigms become invalid more or less

1 Controllable computing elements All parts of

a conventional software system are designed

or adapted (in case of reusing existing

compo-nents) to work together for a common purpose

There is no independent interest or decision of

a single part other than what the system assigns

to it To develop such a system, it is convenient

to decompose the application logic into the

computing elements, in the form of, say,

pro-cedures or objects, and glue up these elements

with predefined coordination mechanisms such

as procedure calls or object method

invoca-tions In open environments, systems are often

built on third-party software entities, i.e.,

var-ious resources and services, which may serve

multiple applications simultaneously and have

their own interests and decisions Although

these entities are, to certain extent, willing to

collaborate, their autonomy must be respected.

In this case the focus of application

develop-ment has to be moved from computation to

coordination, and traditional predefined

coor-dination mechanisms may not suffice

2 Static environment settings Conventional

soft-ware systems aiming at traditional

environ-ment settings are typically developed on the

assumptions of invariant resources, such as

available CPU, memory, underlying system

calls, network bandwidth, and functionalities

provided by other software systems It would

be feasible to make these assumptions when the

environment was generally controllable Such

assumptions no longer hold for systems that

rely on dynamic service integration over the

open Internet environment These systems are

expected to be situational, i.e., capable of

prob-ing and understandprob-ing their runtime contexts

and scenarios, and dynamically evolvable, i.e.,

capable of adapt their structures and behaviors

at runtime to the changes in the environment

they are situated in and the requirements must

satisfy [3]

3 Predefined behavior Conventional software

development requires that delivered ware products exhibit behaviors exactlyconforming to their predefined specifications.However, in the open Internet environment,autonomous entities in the system maydepend on each other, and thus can exhibitun-designed behaviors or unpredicted effects.Conventional software paradigms have noroom to accommodate, or even exploit, such

soft-emergences.

4 Directly verifiable trustworthiness. Thedependability of conventional software sys-tems can be verified and validated usingwell-defined methods and techniques, such

as testing, proving, and reviewing, providedthat sufficient and reliable information aboutthe software and its development process isavailable This availability of informationimplies that the participants unconditionally

trust each other However, software systems

in open environments may involve much moreparticipants than traditional systems becausethey work in a truly decentralized environ-ment The independent interests of differentparticipants and the dynamic environmentmake the trust relationships among themcomplicated and dynamic In this situation,the dependability of a system should be highlyaffected by the trust relationships among theparticipants

There already exist some related observationsand proposals in the literature, e.g System-of-Systems [17], Open Resource Coalitions [18],and Cyber-Physical Systems [19], but a newsoftware paradigm that systematically addressesthese requirements of the new applicationstyles in open network environment is stillmissing

1.2.2 Methodological Reflections

Generally speaking, software development is amental process of devising a solution for a givenproblem, and it must be able to efficiently execute

6 1 Internetware: A Shift of Software …

on a given computing platform Observing

soft-ware paradigms and paradigm shifts in the past

decades, we can see a common scheme of

success-ful paradigms—good balance of the forces from

the problem space, the mental space, and the

com-puting space

The emergence of the structured paradigm in

the 1970s was driven by the rapid development

of the computing power and the need to alleviate

software crises At that time, the normalization of

programming methods and language constructs

to “structured” ones satisfied the need to

man-age the complexity of programming in the mental

space with the now-affordable potential

perfor-mance costs in the computing space, which well

balanced the forces from the mental space and the

computing space

The Object-Oriented paradigm, maturing in the

1980s, is still the mainstream practice in today’s

software industry As the scale and complexity of

application problems increased continuously, the

mental process of software development became

more and more complicated with the regular but

somewhat simplistic constructs of the structured

paradigm, in spite of the advances in

modular-ity, separation of concerns, information hidden,

etc The object-oriented paradigm brought a

soft-ware structuring model well corresponding to the

structure of the problem to be solved, which

sig-nificantly eased the mental process Moreover, the

inheritance, polymorphism, and dynamic binding

mechanisms implemented the mental process of

classification and deduction efficiently in

comput-ing platforms Thus a better balance of the forces

from the problem space, the mental space, and the

computing space was achieved again

Based on the balance of forces from the three

spaces, a software paradigm would provide a

col-lection of tangible artifacts that can be directly

used by developers to regulate their requirements,

design their architectures, and implement their

systems Particularly, the following artifacts are

necessary:

1 Structuring model At the kernel of a software

paradigm, there is a pattern about how

sys-tems are constructed from various software

ele-ments, i.e., what the form of computing

compo-nents is, what kind of coordination mechanismsare used to gluing up these components, andwhy desired system behavior can be achievedwith these computing and coordination

2 Mechanism for flexibility As the “soft” part of

computing, software is always expected to beflexible to certain extent There must be somemechanism to support controlled adjustment ofthe behavior and/or the structure of the sys-tem to deal with the inevitable evolution ofrequirements For example, in structured par-adigm there are procedure encapsulation anddata abstraction to facilitate replacements ofprocedures and data representations, whereas

in the object-oriented paradigm there are itance and dynamic binding to support classextensions and polymorphism

inher-3 Dependability assurance As one of the most

complicated types of artifacts, software tems are prone to the loss of dependability Asoftware paradigm has a collection of associ-ated mechanisms to systematically ensure sys-tem dependability Generally, a set of depend-ability properties and metrics are first specified.Then, some validation and verification tech-niques are used to evaluate these properties.Finally, software improvement processes arecarried out accordingly The last two phases areiterated to keep the software system depend-able in a specific context

sys-4 Supporting tools and environments Reifying

the models and mechanisms above, a set of guages, tools, and integrated environments areneeded to help the developers follow the para-digm efficiently and correctly

lan-Collectively, these artifacts should answer thefour fundamental questions in software develop-ment in a coherent manner: WHAT-IS (what is

to be constructed and executed), HOW-TO-DO(how to develop the required software applica-tions including their constituent entities), HOW-TO-RUN (how to run the software applicationsincluding their constituent entities) and HOW-WELL (how well the constructed and executedsoftware applications can perform)

Last but not least, software paradigm shifts

in the last decades showed that a successful

1.2 The Internetware Paradigm 7

paradigm needs to subsume and be

compati-ble with existing mainstream paradigms, for

social, psychological, economical, and technical

reasons

1.2.3 Architecting Principles

for Internetware

The envisioned Internetware paradigm also needs

to balance the forces from the three spaces

How-ever, the forces are different from those of the

object-oriented paradigm or the structured

par-adigm Specifically, in the problem space, the

trend is to move from solving specific problems

in controlled environments to addressing flexibly

scoped problems with the aim of satisfying

differ-ent users in open environmdiffer-ents The focus is

shift-ing from direct problem solvshift-ing to resource

shar-ing and coordination In the mental space,

mono-lithic view of individual developers or closely

cou-pled developer teams is upgraded to a

service-oriented view of multiple independent developers,

and the focus is shifting from the computing logic

on a specific computer for specific requirements

to the coordination logic on a ubiquitous

com-puting platform with autonomous services

pro-vided for requirements of an open group of users

In the computing space, the computing platform

has evolved from a relatively closed and

control-lable base of standalone or networked computers

to a ubiquitous computing platform in the Internet

environment, and the focus is shifting from

pro-viding computing capabilities to enable and

man-age interactions

In addition to the methodological

considera-tions discussed above, we identify the following

architecting principles for the Internetware:

1 Programmable Coordination As discussed

above, Internetware applications need to be

built on autonomous resources and services

that are provided “AS IS” by independent

owners Mechanisms should be provided to

federate these autonomous software entities

to satisfy specific demands, in a flexible but

disciplined way Thus the coordination aspect

of programming must be explicitly addressed

[20] Beyond providing the communicationchannel between services and adapting theirinterfaces, the mechanism should also supportthe explicit management of the interactionbetween the services, according to the appli-cation logic Furthermore, to adapt to thechanging environment and requirements, thecoordination logic needs to be dynamicallyadaptable, or even evolvable at runtime

2 Situated in open environments Since

Internet-ware application systems are working in anopen, dynamic, and autonomous network envi-ronment, environmental issues are much moredeeply involved in the construction of a sys-tem for Internetware than for conventional par-adigms On the one hand, the system must beable to deal with the changes of the environ-ment at runtime; on the other hand, every par-ticipating component of the system has its ownconcerns on the computing environment andhas to adapt to it accordingly So the Inter-netware paradigm must explicitly support theprobing, understanding, and reconciliation ofenvironment and environment changes Fur-thermore, the autonomy of the participatingcomponents and the locality of the environment

of these components must be respected, whichmeans that simply taking the perceived envi-ronment as system inputs may not work

3 Scalable with multi-level homeostasis To

keep its service satisfactory, the system needs

to adapt to changes of the environment.Thus, software homeostasis mechanisms arenecessary to monitor system behavior anddynamically modify the system to repairabnormalities, or deviations from the expectedbehavior [18] For Internetware, however,the homeostasis mechanism is more than asingle monolithic close-loop control, whichhas limited scalability Since an Internet-ware system is based on the coordinatedcollaboration of a group of autonomouscomponents, there should be a multi-levelmechanism that distributes the responsibility

of homeostasis maintenance to differentlevels (e.g system level, subsystem level,and component level) and different aspects ofsystem (e.g functionality, performance, and

8 1 Internetware: A Shift of Software …

dependability) according to their capabilities

and responsibilities in the system

4 Dependable with explicit trust management.

To ensure the dependability of a system

in open environments, it is unavoidable to

deal with the complicated trust relationships

among multiple independent participants, such

as primitive service providers, value-adding

service providers, brokers, and end users Trust

management mechanisms, which are

origi-nally proposed for decentralized authorization

and authentication, could be brought to deal

with the complicated trust relationships among

participants Combining trust management

and conventional verification and verification

techniques, a framework for the dependability

assurance of Internetware applications should

be provided

1.2.4 Our Roadmap and Explorations

There is a long way towards a full-fledged

Internetware paradigm To achieve the final goal

of shaping the complete methodology, building

the comprehensive supporting environment and

developing killer applications of Internetware,

we need a stepwise roadmap in which a set

of models, techniques, supporting facilities,

and applications can be gradually developed

following the identified principles (Fig.1.1)

To this end, we first propose an open dination model by extending the conventionalobject-oriented paradigm for a flexible andadaptive structuring technique With the cor-responding middleware, it demonstrates thecoordinated application in open environment andlays the groundwork for Internetware research

coor-A succeeding environment-driven model sizes the concerns about the coordinated softwaresystems and the environment where they actuallysituate in and adapt to Environment facilities arealso developed to sense, monitor, and manage theenvironment Cooperating with the mechanismsprovided in the coordination middleware, theenvironment-interacting adaptive software sys-tems can thus be realized The dependability ofthe system is finally assured with a framework,which has certain trust management models,mechanisms, and facilitates to turn the systemsinto trusted ones Although the three levels ofthe work are still preliminary, we hope they canillustrate our vision of the Internetware

synthe-1.3 Open Coordination Model

1.3.1 Coordination Model

for Internetware

The construction of Internetware software tems needs to primarily coordinate the existing

sys-Fig 1.1 Our roadmap and

efforts toward Internetware Open CoordinaƟon

Model

Environment DrivenModelDependability

CoordinaƟon Middleware

Environment FaciliƟes

Trust Management

Coordinated ApplicaƟons

AdapƟve ApplicaƟons

Trusted ApplicaƟons

InternetwareMethodology

SupporƟng Environment

Killer ApplicaƟons

1.3 Open Coordination Model 9

software entities and resources over the Internet

By the term of coordination model, it basically

refers to the specification of inter-relationships

among the computing entities and the regulations

between their interaction [21] To follow the

prin-ciples and meet the challenges discussed in the

previous section, a coordination model for

Inter-netware is required to be in a position to

accom-modate to the high dynamism in the systems and

facilitate the possible changes, either foreseeable

or unforeseeable, to take place At the same time,

it should provide the developers and users a good

understanding for the systems during the

life-cycle of the systems By such means, the

com-plexities in the systems can be well controlled

There have been various coordination models

proposed and they are briefly categorized as

data-driven and control-driven [20] Although

the data-driven models, which usually advocate

some shared data spaces, have be argued to be

fit for the open systems [22], they have inherent

short-comings, such as their insufficiency in

isolating the computing and coordinating logic

and indigestibility in specifying the system

structure In the models of the latter category,

software entities, or say components, have well

defined interfaces The coordination is driven by

propagating the state changes among components

with some interacting mechanisms

The object-oriented programming model

introduces the classic control-driven coordination

schema with the form of “O.m()” This object

granularity model is also extended into the similar

ones for the components [20,23] and services

[24] However, the core of the coordination is

always reserved, which is to establish a fixed

linkage between the objects with specific

mes-saging facility It is not difficult to see that this

schema fits itself quite well in a closed computing

environment with idealized assumptions, as

argued before As long as the computing entities

are all under the control, the system can be

modeled by this means and executed with no

problem

Nevertheless, the mechanism like “O.m()”

consolidates the original coordination motivation

and environment assumption [25] Once thesefactors change, which are assumed to happenrarely, major re-organization of the coordinationcan be caused However, the open environment

of Internet breaks down those assumptions Thedynamism lying in all aspects of the systemsituated in such an environment needs to be fullyconsidered into the construction of the system

so that the changes can be effectively handled.Structurally, the inter-relationships of the entitieswithin coordination could be more complicatedthan referring to each other directly and fixedly,and these relationships have to be determined

on the runtime factors, such as availability andprovision status of the software entities And, theonefold interaction mechanism is also required

to be enriched Effective and dependable tions should be achieved to respect the dynamicnetwork environment where the software entitiesare situated and the autonomous interactingschema that they prefer

OO coordination mechanism with flexibilityand adaptability, such as polymorphism [26]and some advanced communicating techniques[27,28], respectively But ideally, the systems thatare hosted in an open environment still need to

be modeled with a unified model From the study

OO schema, the model should take an abstractand indirect way to shape the coordinating sys-tems and organize the interacting entities Thusthe changes can be applied at some relativelyhigh level of the model to avoid the massivemodifications in the low level relations Also, theinteractions should also be modeled explicitly

to enable the customization on the runtimecommunication among the coordinated entities.The actual execution of the system should bedetermined at runtime by investigating the onsiteconfiguration of the coordination Also, there areother conceivable advanced requirements on the

10 1 Internetware: A Shift of Software …

coordination, such as the model level analysis

and evaluation for the system non-functional

qualities

Overall, the coordination model as a kernel

of the Internetware systems should carry enough

information to put the system into execution,

and more importantly, to get the adaptation of

the system into effective operation Beyond

the conceptional modeling, a coordination

sup-porting platform also needs the mechanisms to

realize the potential adaptability brought by the

model as a whole structuring of Internetware

systems

1.3.2 A Software Architecture Based

Approach

Toward an effective open coordination model for

the structuring of Internetware systems discussed

in the previous section, we proposed a modeling

approach based on software architecture To put

the coordination into execution, an intrinsic model

implementation is introduced and adaptation

facil-itates are also advised

Coordination Model

Software architecture “involves the description

of elements from which systems are built,

interactions among those elements, patterns that

guide their composition, and constraints on these

patterns” [29] It is originally viewed as design

specification and recently also employed in the

research of software coordination [30–33] With

software architecture, the system components are

organized and coordinated with connectors as

inter joints It offers explicit and programmable

coordination logic Especially, the connectors

are treated as first class programming concepts

This allows connectors to be reconfigured

dynamically to reflect coordination changes in

addition to component adaptations Software

architecture supports the early analysis of system

non-functional properties, which is one of its

motivating benefits Also there are various formal

models and verification tools proposed to help the

ensuring of architectural correctness Software

architecture embodies essential information thatshould be held by the implementation to supportand regulate future adaptation and evolution.These benefits make software architectureaccepted as a relative ideal model for softwarecoordination

Implementation

To implement the coordination with the model

of software architecture, techniques such asarchitecture refinement and explicit architecturemaintenance, to name a few, are developed.These existing approaches share the issuethat difficult and ad-hoc efforts are needed tomaintain the consistency between a softwarearchitecture specification and the working systemimplementation

Thereby, we propose an intrinsic approach

to actually implement the coordination thatsoftware architecture models The intrinsicmechanism fully integrates coordination into theprogramming model The software architecture

is encoded into some data structure and explicitlyreified as a first class object, namely an archi-tecture object, in the final implementation Therunning components are glued with this object

by delegating the runtime interpretation of objectreferences and associated method invocations inthe form of “O.m()” to the architecture object Inother words, the references and invocations are

“functions” over the current software architectureconfiguration By this means, the coordinationconcerns are separated from interacting compo-nent objects The final coordination being carriedout complies with the specified model of softwarearchitecture as the reified software architecture(which is implementation And any change of thearchitecture object, in both the components andconnectors, will immediately affect the runtimecoordination process

Adaptations

Based on the intrinsic coordination tation, it is natural to realize dynamic systemadaptation as the reconfiguration behavior of thearchitecture object Above the primitive adap-

implemen-1.3 Open Coordination Model 11

tation actions, e.g., component add/remove and

connector replacement, more sophisticated ones

can be defined as the methods of the

architec-ture object, by which means the concrete

adap-tion process is specified for that type of the

archi-tecture The implementation of these methods is

mainly changing the topology with the facilities

provided in base architecture class

Also, there could be some system

reconfigura-tion requirements gradually discovered after the

system was put into operation Common solutions

for these unanticipated reconfigurations require

a system restart However, the stop of service

could bring a high cost in some circumstance,

esp when there were valuable data not persistently

stored Our approach also provides a reasonable

support for unanticipated dynamic

reconfigura-tions A new subclass of the original architecture

class should be defined to implement new

recon-figuration behavior The consistency between the

new and old architectures is ensured with the type

consistence checking against the classes defining

those architectures

Summarily, the proposed software

architec-ture based coordination model gives a unified

approach to specify the interrelation of the

com-ponents as well the interactions among them

The featured intrinsic implementation with the

corresponding adaptation mechanisms provides

the basis of the flexible and adaptable system

structuring technique for Internetware systems

Besides, we also introduce an architecture

check-ing technique based on graph-grammar [34] A

coordination middleware is also designed and

implemented with some running cases All these

the works serve as the first step toward

Internet-ware

1.4 Environment-Driven Model

1.4.1 Structure and Dynamics

Internetware application systems run in the

open, dynamic, and uncertain environment

For such a system, the runtime adaptation

to frequent changes in the environment and

users’ requirements must be supported as an

indispensable feature rather than handled asexceptions Traditional software systems running

in the principally closed, static, and controllableenvironment usually have their processing of theenvironmental information implicitly consoli-dated into the application logic Such an approachcannot effectively cope with the dynamics of theopen environment

To adapt themselves at runtime in anenvironment-driven manner, Internetware appli-cation systems need to process the environmentinformation explicitly and separately Basicallythe environment context needs to be explicitlycollected, understood, and utilized However,

if we monolithically divide the entire systeminto coordination of constituent componentsand explicit environment processing facilities,scalability of the system will be quite limited

We cannot effectively cope with the complexity

of implementing, deploying, and operatingInternetware application systems in the openenvironment This is mainly because Internetwaresystems are the coalition of autonomous services

in a large and fully-distributed environment.Such systems are commonly multi-level onesconstructed from subsystems, and may have thefractal or self-similar characteristics Differentconstituent components of the system have theirown concerns of the computing environment andhave to adapt to the environment independently.Take the traffic control system as an example

In the district level, software entities for collectingand disseminating traffic information coordinatewith each other to inform the drivers of the trafficinformation and instruct them to avoid the traf-fic jam in each district of the city Traffic controlsystems in the district level coordinate with eachother and form the traffic control system in the citylevel Such system can further coordinate to formsystems in upper levels of larger scale The trafficcontrol system in each level has similar structure.Since the sub-systems in each level has theirown concerns of the environment context, theenvironment processing facilities for Internetwareapplication systems cannot be a monolithic soft-ware entity It must also have the fractal charac-teristics, in accordance with the multi-level coor-dination of the Internetware application systems

12 1 Internetware: A Shift of Software …

System Structure

One constituent part of an Internetware

applica-tion system consists of three parts:

• Goal-driven adaptation logic Goal of the users

specifies expected behavior of the coalition

in given situations The goal-driven

adapta-tion logic decides how to change the

coordina-tion in different situacoordina-tions A unified semantic

framework for users’ goal and the environment

processing facilities is required for the logic

reasoning Customizable adaptation rules and

reification mechanisms are also necessary for

realizing decisions of the adaption logic

• Open coordination subsystem The open

coor-dination subsystem is responsible for the

busi-ness logic of the application It also supports

the dynamic reconfiguration of itself, in order

to smoothly carry out the adaptation, as

dis-cussed in Sect.1.3

• Environment processing facilities Expected

behavior of the coalition corresponds to

situ-ation of the environment, which necessitates

the explicit environment processing facilities

Such facilities are in charge of collecting,

understanding, and utilizing the environmentinformation They transform raw environmentdata to high-level representations of situationsfor the coalition

The structure of the environment processingfacilities has the fractal or self-similar character-istics, as shown in Fig.1.2 For a specific part ofthe system situated in the environment, it has theenvironment-driven software structure discussedabove Subsystems further coordinate to form sys-tems in a higher level, and the coordination in eachlevel has similar software structure We argue thatsuch a fractal structure of environment processingfacilities is a promising approach to achieve sys-tem scalability and cope with system complexityissues for Internetware application systems in theopen environment

Homeostasis

From the behavioral perspective, an Internetwareapplication system can be viewed as a homeosta-sis of the coalition among services Homeostasis isthe propensity of a system to automatically restoreits desired equilibrium state when changes in the

Open Environment Environment

Handling FaciliƟes Open CoordinaƟon Subsystem

Goal-driven AdaptaƟon Logic