116,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Sprache: Englisch
Understand the computing technology that will power a connected future
The explosive growth of the Internet of Things (IoT) in recent years has revolutionized virtually every area of technology. It has also driven a drastically increased demand for computing power, as traditional cloud computing proved insufficient in terms of bandwidth, latency, and privacy. Edge computing, in which data is processed at the edge of the network, closer to where it’s generated, has emerged as an alternative which meets the new data needs of an increasingly connected world.
Edge Computing offers a thorough but accessible overview of this cutting-edge technology. Beginning with the fundamentals of edge computing, including its history, key characteristics, and use cases, it describes the architecture and infrastructure of edge computing and the hardware that enables it. The book also explores edge intelligence, where artificial intelligence is integrated into edge computing to enable smaller, faster, and more autonomous decision-making. The result is an essential tool for any researcher looking to understand this increasingly ubiquitous method for processing data.
Edge Computing readers will also find:
- Real-world applications and case studies drawn from industries including healthcare and urban development
- Detailed discussion of topics including latency, security, privacy, and scalability
- A concluding summary of key findings and a look forward at an evolving computing landscape
Edge Computing is ideal for students, professionals, and enthusiasts looking to understand one of technology’s most exciting new paradigms.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 469
Veröffentlichungsjahr: 2025
Ähnliche
Table of Contents
Cover
Table of Contents
Title Page
Copyright
About the Authors
Preface
About the Companion Website
1 Why Do We Need Edge Computing?
1.1 The Background of the Emergence
1.2 The Evolutionary History
1.3 What Is Edge Computing?
1.4 Summary and Practice
Chapter 1 Suggested Papers
References
2 Fundamentals of Edge Computing
2.1 Distributed Computing
2.2 The Basic Concept and Key Characteristics of Edge Computing
2.3 Edge Computing vs. Cloud Computing
2.4 Summary and Practice
Chapter 2 Suggested Papers
References
3 Architecture and Components of Edge Computing
*
3.1 Edge Infrastructure
3.2 Edge Computing Models
3.3 Networking in Edge Computing
3.4 Summary and Practice
Chapter 3 Suggested Papers
References
Note
4 Toward Edge Intelligence
*
4.1 What Is Edge Intelligence?
4.2 Hardware and Software Support
4.3 Technologies Enabling Edge Intelligence
4.4 Edge Intelligent System Design and Optimization
4.5 Summary and Practice
Chapter 4 Suggested Papers
References
Note
5 Challenges and Solutions in Edge Computing
*
5.1 Programmability and Data Management
5.2 Resource Allocation and Optimization
5.3 Security, Privacy, and Service Management
5.4 Deployment Strategies and Integration
5.5 Foundations and Business Models
5.6 Summary and Practice
Chapter 5 Suggested Papers
References
Note
6 Future Trends and Emerging Technologies
*
6.1 Edge Computing and New Paradigm
6.2 Integration with Artificial Intelligence
6.3 6G and Edge Computing
6.4 Edge Computing in Space Exploration
6.5 Summary and Practice
Chapter 6 Suggested Papers
References
Note
7 Case Studies and Practical Applications
*
7.1 Manufacturing
7.2 Telecommunications
7.3 Healthcare
7.4 Smart Cities
7.5 Internet of Things
7.6 Retail
7.7 Autonomous Vehicles
7.8 Summary and Practice
Chapter 7 Suggested Papers
References
Note
8 Privacy and Bias in Edge Computing
*
8.1 Privacy in Edge Computing
8.2 Accessibility and Digital Divide
8.3 Summary and Practice
Chapter 8 Suggested Papers
References
Note
9 Conclusion and Future Directions
9.1 Key Insights and Conclusions
9.2 So, What Is Next?
Index
End User License Agreement
List of Tables
Chapter 2
Table 2.1 Edge computing vs. cloud computing.
Chapter 4
Table 4.1 Comparison of hardware types for edge AI applications: performance...
Chapter 5
Table 5.1 Technical challenges in edge computing.
Chapter 6
Table 6.1 Relationship of different computing paradigms to edge computing.
Table 6.2 Model parameter scales.
Table 6.3 Comparison of satellites in different orbit.
Chapter 8
Table 8.1 Original dataset.
Table 8.2 Anonymized dataset with 2‐anonymity.
List of Illustrations
Chapter 1
Figure 1.1 The increase of data pushes the evolution of edge computing.
Figure 1.2 Number of publications searched by keyword “edge computing” and “...
Figure 1.3 The evolution of edge computing and key milestones.
Figure 1.4 “Edge computing” and “fog computing” trends.
Figure 1.5 Edge computing paradigm.
Figure 1.6 Edge computing is a continuum.
Figure 1.7 “Edge computing” and “edge AI” trends.
Chapter 2
Figure 2.1 Traditional cloud computing model.
Figure 2.2 Edge computing model characteristics.
Figure 2.3 Round trip time between client and edge/cloud.
Figure 2.4 Bandwidth between client and edge/cloud.
Figure 2.5 Round trip time for processing audio command on edge and cloud.
Chapter 3
Figure 3.1 Three layers of edge.
Figure 3.2 “MEC” and “cloudlet computing.”
Figure 3.3 Edge‐to‐edge collaboration.
Figure 3.4 Edge‐to‐device collaboration.
Figure 3.5 Edge‐to‐cloud collaboration.
Figure 3.6 Cloud‐edge‐device collaboration.
Figure 3.7 Edge computing and networking.
Figure 3.8 The architectural framework for edge computing‐network integratio...
Chapter 4
Figure 4.1 Motivation of edge intelligence.
Figure 4.2 Dataflow of edge intelligence.
Figure 4.3 Edge intelligence.
Figure 4.4 TPU.
Figure 4.5 VPU.
Figure 4.6 Jetson Nano.
Figure 4.7 TrueNorth chip.
Figure 4.8 FPGA.
Figure 4.9 Nvidia AGX Xavier.
Figure 4.10 Container.
Figure 4.11 Overview of pruning.
Figure 4.12 Overview of quantization.
Figure 4.13 Overview of quantization‐aware training (QAT).
Figure 4.14 Overview of post‐training quantization (PTQ).
Figure 4.15 Overview of knowledge distillation.
Figure 4.16 Overview of hardware‐software codesign.
Figure 4.17 The framework of CLONE. The CLONE framework operates by allowing...
Figure 4.18 An overview of the TensorRT‐enabled framework, which integrates ...
Figure 4.19 A depiction of the collaborative inference pipeline for the mult...
Chapter 5
Figure 5.1 Edge computing paradigm.
Figure 5.2 The naming mechanism for the edge operating system (edgeOS).
Figure 5.3 Data abstraction in edge computing scenarios.
Figure 5.4 Offloading framework in edge computing scenarios for vehicle‐edge...
Figure 5.5 The design of ChatCache.
Figure 5.6 Storage system architecture.
Figure 5.7 An example of function consolidation and deduplication. Each edge...
Chapter 6
Figure 6.1 Possible sky computing architecture.
Figure 6.2 Problems of existing computing paradigms.
Figure 6.3 Problems of existing computing paradigms.
Figure 6.4 A hierarchical network slicing architecture.
Figure 6.5 Architecture of multi‐access edge learning‐based offloading (MELO...
Figure 6.6 Kodan architecture design.
Chapter 7
Figure 7.1 A taxonomy of edge computing applications.
Figure 7.2 Edge computing in manufacturing.
Figure 7.3 Edge computing in telecommunications.
Figure 7.4 Edge computing in healthcare.
Figure 7.5 High‐level view of an IoT‐based smart city.
Figure 7.6 High‐level view of an IoT‐based smart retail.
Figure 7.7 High‐level view of edge computing for autonomous vehicles.
Chapter 8
Figure 8.1 Privacy forms.
Figure 8.2 An example of homomorphic encryption.
Figure 8.3 Federated learning.
Guide
Cover
Table of Contents
Title Page
Copyright
About the Authors
Preface
About the Companion Website
Begin Reading
Index
End User License Agreement
Pages
ii
iii
iv
xiii
xv
xvi
xvii
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
IEEE Press
445 Hoes LanePiscataway, NJ 08854
IEEE Press Editorial Board
Sarah Spurgeon,
Editor‐in‐Chief
Moeness Amin
Jón Atli Benediktsson
Adam Drobot
James Duncan
Hugo Enrique Hernandez Figueroa
Ekram Hossain
Brian Johnson
Hai Li
James Lyke
Joydeep Mitra
Albert Wang
Desineni Subbaram Naidu
Yi Qian
Tony Quek
Behzad Razavi
Thomas Robertazzi
Patrick Chik Yue
Edge Computing
Systems and Applications
Lanyu XuDepartment of Computer Science and EngineeringOakland University, RochesterMichigan, United States
Weisong ShiDepartment of Computer and Information SciencesUniversity of Delaware, NewarkDelaware, United States
Copyright © 2025 by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved.
Published by John Wiley & Sons, Inc., Hoboken, New Jersey.Published simultaneously in Canada.
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, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per‐copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750‐8400, fax (978) 750‐4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748‐6011, fax (201) 748‐6008, or online at http://www.wiley.com/go/permission.
The manufacturer's authorized representative according to the EU General Product Safety Regulation is Wiley‐VCH GmbH, Boschstr. 12, 69469 Weinheim, Germany, e‐mail: [email protected].
Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional 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.
For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762‐2974, outside the United States at (317) 572‐3993 or fax (317) 572‐4002.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.
Library of Congress Cataloging‐in‐Publication Data Applied for
Hardback ISBN: 9781394285839
Cover Design: WileyCover Image: © VectorMine/stock.adobe.com
About the Authors
Lanyu Xu is currently an Assistant Professor of Computer Science and Engineering at Oakland University. Dr. Xu leads the Edge Intelligence Systems Lab. Her research intersects edge computing and deep learning, emphasizing the development of efficient edge intelligence systems. Her work explores optimization frameworks, intelligent systems, and AI applications to address challenges in efficiency and real‐world applicability of edge systems across various domains.
Weisong Shi is an Alumni Distinguished Professor and Department Chair of Computer and Information Sciences at the University of Delaware (UD). Dr. Shi leads the Connected and Autonomous Research (CAR) Laboratory. He is an internationally renowned expert in edge computing, autonomous driving, and connected health. His pioneer paper, “Edge Computing: Vision and Challenges,” has been cited over 8000 times. He is the Editor‐in‐Chief of IEEE Internet Computing Magazine and the founding steering committee chair of several conferences, including the ACM/IEEE Symposium on Edge Computing (SEC), IEEE/ACM International Conference on Connected Health (CHASE), and the IEEE International Conference on Mobility (MOST). He is a fellow of IEEE and a distinguished member of Association for Computing Machinery (ACM).
Preface
Over the past decade, edge computing's rapid evolution has fundamentally transformed how data is processed, stored, and utilized across multiple industry sectors, such as smart manufacturing, healthcare, smart cities, and transportation. As a critical enabler of technologies such as the Internet of Things (IoT), autonomous systems, and real‐time analytics, edge computing has progressed from a nascent concept to widespread adoption. Despite this remarkable growth, there remains a lack of educational resources dedicated to equipping the next generation of the workforce with the knowledge and skills needed to advance edge computing further.
This book was motivated to address that gap, offering a comprehensive introduction to edge computing's principles, architectures, applications, and challenges. It aims to provide readers, ranging from students to professionals, with a solid foundation in edge computing, enabling them to understand its current state, tackle its challenges, and drive its development. By bridging theory and practice, this book aspires to inspire innovation, foster collaboration, and promote growth in this rapidly evolving field.
Designed as both a textbook and a reference guide, this book includes practice questions, course projects, and curated reading materials for each chapter to enhance learning. Readers with diverse interests and goals can navigate directly to the chapters most relevant to them, making the book a flexible resource for students, educators, researchers, and professionals alike.
The book is structured into nine chapters. Chapter 1 introduces the importance of edge computing, providing its background and evolutionary history. Chapter 2 lays the groundwork, covering fundamental principles, models, and technologies that underpin edge computing. Chapter 3 delves into the architecture and components of edge computing, including infrastructure and collaborative models. Chapter 4 transitions into edge intelligence by highlighting the integration of artificial intelligence with edge computing. Chapter 5 addresses key challenges such as programmability, resource optimization, and security, while proposing potential solutions. Chapter 6 looks to the future, discussing emerging paradigms like sky computing, 6G, and edge computing in space exploration. Chapter 7 provides practical insights through real‐world case studies, illustrating edge computing's impact on industries such as manufacturing, healthcare, smart cities, and more. Chapter 8 examines privacy concerns and the digital divide, exploring biases, their impacts, and mitigation strategies in edge computing. Chapter 9 concludes the book.
This endeavor would not have been possible without the unwavering dedication and expertise of the team behind this book. We are deeply grateful to the contributors, editors, and reviewers whose insights and hard work shaped this book. In particular, we extend our heartfelt thanks to Dr. Haihua Chen (University of North Texas), Dr. Shihong Hu (Hohai University), Dr. Sidi Lu (William & Marry), Dr. Kewei Sha (University of North Texas), Dr. Qingyang Zhang (Anhui University), Dr. Xingzhou Zhang (Chinese Academy of Science), and PhD students Komala Subramanyam Cherukuri (University of North Texas), Yuankai He (University of Delaware), Shaibal Saha (Oakland University), Qiren Wang (University of Delaware), Yichen Xia (University of Delaware), and Yongtao Yao (University of Delaware).
To all the readers, we hope you enjoy reading the book and find the book serves as both a resource and an inspiration as you explore the exciting world of edge computing.
January, 2025
Lanyu Xu
Rochester, United States
Weisong Shi
Newark, United States
About the Companion Website
This book is accompanied by a companion website:
www.wiley.com/go/xu/computing
This website contains the PowerPoint slides for each chapter.
1Why Do We Need Edge Computing?
What is edge computing? Why did it become popular after being proposed? What are the relationships between edge computing and IoT/Cloud Computing? In this chapter, we will answer these three questions by introducing the background, the evolutionary history, and the concept of edge computing.
1.1 The Background of the Emergence
To answer the question of this chapter, let us trace back to when edge computing was proposed, back to the big data era when the Internet of Things (IoT) and cloud computing were blooming.
The IoT technology [3] aims to connect physical objects to the Internet according to the communication protocols of IoT, utilizing technologies such as RFID (radio frequency identification), wireless data communication, and GPS (global positioning system). This enables information exchange for intelligent identification, positioning, tracking, monitoring, and management of Internet resources. IoT has significantly expanded with the advancement of computer and network communication technologies. It now encompasses the integration of almost all information technologies with computer and network technologies, facilitating real‐time data sharing between objects and achieving intelligent real‐time data collection, transmission, processing, and execution. The concept of “computer information perception without human intervention” has gradually been applied to fields such as wearable devices, smart homes, environmental sensing, intelligent transportation systems, and smart manufacturing [18, 36]. Key technologies involved in IoT include:
Sensor Technology
: This involves acquiring information from natural sources, processing (transforming), and identifying it. Sensor technology is a critical aspect of computer applications, as it senses (or responds to) and detects specific information from the measured object, converting it into output signals according to certain rules.
RFID Technology
: This comprehensive technology integrates radio frequency and embedded technologies to automatically identify target objects and obtain related data through radio frequency signals. The identification process does not require human intervention and can operate in various harsh environments, with promising and broad applications in automatic identification, logistics management, and more.
Embedded System Technology
: This is a complex technology that integrates computer hardware and software, sensor technology, integrated circuit technology, and electronic application technology. Over the decades, intelligent terminal products characterized by embedded systems have become ubiquitous, ranging from smartwatches to aerospace satellite systems. Embedded systems are transforming people's lives, driving industrial production, and advancing the defense industry. If we make a simple analogy of the IoT to the human body, sensors are akin to human senses like eyes, nose, and skin; the network is the nervous system transmitting information, and the embedded system is the brain that classifies and processes the received information.
Later on, with the rapid development of IoT and the widespread adoption of 4G/5G wireless networks, the era of the Internet of Everything (IoE) [11] has arrived. Cisco introduced the concept of IoE in December 2012. It represents a new network architecture for future Internet connectivity and the evolution of IoT, enhancing the network's intelligent processing and security features. IoE employs a distributed structure, integrating application‐centric networks, computing, and storage on a new platform. It is driven by IP settings, global higher bandwidth access, and IPv6, supporting hundreds of millions of edge terminals and devices connected to the Internet. Compared to IoT, IoE not only involves “thing‐to‐thing” connections, but also introduces a higher level of “human‐to‐thing” connectivity. Its distinguishing feature is that any “thing” will possess contextual awareness, enhanced computing capabilities, and sensing abilities.
Integrating humans and information into the Internet, the network will have billions or even trillions of connected nodes. The IoE is built on the physical network, enhancing network intelligence to achieve integration, coordination, and personalization among the “things” on the internet.
Application services based on the IoE platform require shorter response times and will generate a large amount of data involving personal privacy. For example, sensors and cameras installed on autonomous vehicles capture road condition information in real time; one car with five cameras can generate more than 24 terabytes (TB) data per day [17]. According to the Insurance Institute for Highway Safety, there will be 3.5 million self‐driving vehicles on U.S. roads by 2025 and 4.5 million by 2030 [21]. The Boeing‐787 generates about 5 gigabytes (GB) of data per second and requires real‐time processing of the data. In Beijing, China, the electric vehicle monitoring platform can provide continuous ‐hour real‐time monitoring for 10,000 electric vehicles and forward data to various enterprise platforms at a rate of one data point every 10 seconds per vehicle. In terms of social security, the United States has deployed over 30 million surveillance cameras, generating more than 4 billion hours of video data each week. China's “Skynet” surveillance network, used for crime prevention, has installed over 20 million high‐definition surveillance cameras nationwide, monitoring and recording pedestrians and vehicles in real time.
Since the concept was proposed in 2005, cloud computing has been widely applied, changing how people work and live. SaaS (Software as a Service) is commonly used in data centers of major IT companies like Google, Twitter, Facebook, and Baidu. Scalable infrastructure and processing engines supporting cloud services have significantly impacted application services such as Google File System (GFS), MapReduce programming model, Hadoop (a distributed system developed by Apache Foundation), and Spark (the in‐memory computing framework designed by the AMP Lab at the University of California Berkeley). However, in the context of IoT and similar applications, data is geographically dispersed and demands higher response times and security. Although cloud computing provides an efficient platform for big data processing, the network bandwidth growth rate cannot keep up with the data growth rate. The cost reduction rate of network bandwidth is much slower than that of hardware resources like CPU and memory, and the complex network environment makes it challenging to significantly improve network latency. Therefore, the traditional cloud computing model will struggle to support application services based on IoE efficiently and in real time, requiring solutions to address the bandwidth and latency bottlenecks.
With the rapid development and widespread application of the IoE, edge devices are transitioning from primarily serving as data consumers to serving as both data producers and consumers. Simultaneously, network edge devices are gradually capable of utilizing the collected real‐time data for pattern recognition, predictive analysis or optimization, and intelligent processing. In the edge computing model, computing resources are closer to the data source, and network edge devices now have sufficient computational power to process the raw data locally and send the results to the cloud computing center locally. The edge computing model not only reduces the bandwidth pressure in network transmission, speeding up data analysis and processing, but also lowers the risk of privacy leaks for sensitive terminal data.
Currently, big data processing is shifting from the centralized processing era centered on cloud computing (we refer to the years from 2005 to 2015 as the centralized big data processing era) to the edge computing era centered on the IoE (we refer to it as the edge‐based big data processing era). During the centralized big data processing era, the focus was more on centralized storage and processing of big data, achieved by building cloud computing centers and leveraging their powerful computing capabilities to solve computational and storage issues centrally. In contrast, in the edge‐based big data processing era, network edge devices generate massive real‐time data. In 2018, Cisco's Global Cloud Index estimated that nearly 850 zettabytes (ZB) will be generated by all people, machines, and things by 2021. Yet only around 10% is classed as useful data; useful data is predicted to four times exceed data center traffic (21 ZB per year) [10]. From 2018 to 2023, the average number of devices owned per person worldwide increased from 2.4 to 3.6. Specifically, in North America, on average, one person owned eight devices in 2018 and 13 devices in 2023 [38]. According to Statista, the number of IoT devices connected to the network was 15.14 billion in 2023 and will reach 29.42 billion in 2030 [35]. This mismatch between data producing and data consuming requires the emergence of an alternation for cloud‐based data centers. Instead of purely relying on cloud computing, data can be stored, processed, and analyzed at the network edge. These edge devices will be deployed on edge computing platforms supporting real‐time data processing, providing users with numerous service or function interfaces, which users can invoke to obtain the necessary edge computing services.
Therefore, the linearly growing centralized cloud computing capacity can no longer match the exponential growth of massive edge data. Single computing resources based on the cloud computing model can no longer meet the demands for real‐time processing, security, and low energy consumption in big data processing. Based on the existing centralized big data processing centered on the cloud computing model, there is an urgent need for edge big data processing technology centered on the edge computing model to handle the vast edge data. The two complement each other, applied to big data processing at both the cloud center and the edge end, addressing the inadequacies of cloud computing services in the IoE era.
When observing the data explosion in three dimensions: velocity, variety, and volume, we will find that the emergence and rapid development of edge computing is inevitable (Figure 1.1). The cloud‐centralized computing paradigm performs well when the data is generated with a confined speed, size, and format. While in the IoE era, data is increasingly produced at the network's edge regarding velocity, variety, and volume. In terms of variety, different types of data (e.g., text, audio, and photo) are generated every day and every second from numerous devices (e.g., IoT, web browser, camera, and social media). These data are generated with different velocities (e.g., real time, near real time, periodic, and batch generated). Therefore, storage and process requirements for these data will be different. With the tremendous number of devices and the frequent speed of data generation, there is no surprise that the volume of data generated is increasing dramatically. Megabytes (MB) and TB have become the typical units. In fact, as of 2024, the amount of data generated per day is around 328.77 million TB, which equals 0.33 ZB [13]. Given these factors, relying purely on the cloud for all data processing is impossible. This is not just because of the computing pressure brought to the cloud computation center but also because the bandwidth capability required for transmitting this amount of data is challenging. The only solution to process the data reliably and in a timely manner is edge computing, which ensures a shorter response time, more efficient processing, and smaller network pressure.
Figure 1.1 The increase of data pushes the evolution of edge computing.
Compared to cloud computing, edge computing can better support mobile computing and IoT applications, offering the following distinct advantages:
Greatly Alleviates Network Bandwidth and Data Center Pressure
: With the development of IoT, global devices will generate massive amounts of data. However, only a small portion of this data is critical, while most of it is temporary and does not need long‐term storage (the amount of data generated by devices is two orders of magnitude higher than the amount of data that needs to be stored). Edge computing can fully utilize geographically distributed network edges to process a large amount of temporary data, thereby reducing the pressure on network bandwidth and data centers.
Enhances Service Responsiveness
: The inherent limitations of mobile devices in computing, storage, and power resources are evident. Cloud computing can provide services to mobile devices to address these deficiencies. However, network transmission speeds are constrained by the development of communication technologies, and in complex network environments, issues such as unstable connections and routing further exacerbate latency, jitter, and slow data transmission speeds, severely affecting the responsiveness of cloud services. Edge computing offers services near the user, ensuring low network latency through proximity and reducing network jitter with more straightforward routing. With the development of 5G and 6G, the diverse application scenarios and differentiated service requirements pose challenges to 5G/6G networks regarding throughput, latency, number of connections, and reliability. Edge computing and 5G/6G technologies complement each other, with edge computing leveraging localization, proximity, and low latency to drive 5G/6G architectural changes, while 5G/6G technology is essential for reducing data transmission latency and enhancing service responsiveness in edge computing systems.
Protects Privacy Data and Enhances Data Security
: Data security has always been a critical issue in IoT applications. Surveys show that approximately 86% of the U.S. general population is concerned about data privacy
[23]
. In the cloud computing model, all data and applications are stored in data centers, making it difficult for users to have fine‐grained control over data access and usage. Edge computing provides the infrastructure for storing and using critical privacy data, restricting the operation of privacy data within firewalls and thereby enhancing data security (for more detailed information, see
Chapter 8
).
1.2 The Evolutionary History
The field of edge computing has developed rapidly since 2014. We categorize the development process into three stages: technology preparation period, rapid growth period, and steady development period. We use “edge computing” as the keyword to search the number of articles published per year in Google Scholar. As shown in Figure 1.2, before 2015, edge computing was in the technology preparation period. Since 2015, the number of papers related to “edge computing” has grown tenfold. Edge computing has entered a rapid growth period. The number of papers has been increasing and reaching a steady development period since 2020. In this period, the development is focused on integrating academia and industry, bringing the product into the business, and finally facilitating peoples' daily lives. Figure 1.3 illustrates typical events in the development process of edge computing.
The development of edge computing is closely linked to the evolution of data‐oriented computing models. As the scale of data increases, the demand for performance and energy efficiency in data processing continues to grow. To address the issues of computational load and data transmission bandwidth in data transfer, computation, and storage processes, researchers explored ways to enhance data‐processing capabilities near data sources even before the advent of edge computing. This involves shifting computational tasks from centralized computing centers to the network edge. The main typical models include distributed database models, peer‐to‐peer (P2P) computing models, content delivery network (CDN) models, mobile edge computing models, fog computing models, and cloud‐sea computing models. We will explain these different models in the order of their emergence and also introduce the history of edge computing.
Figure 1.2 Number of publications searched by keyword “edge computing” and “edge intelligence.”
Figure 1.3 The evolution of edge computing and key milestones.
1.2.1 Technology Preparation Period
During the technology preparation period, edge computing went through the development process of dormancy, presentation, definition, and generalization.
1.2.1.1 Distributed Database Models
The distributed database model results from combining database technology and network technology. In the era of big data, the growth in the variety and quantity of data has made distributed databases a core technology for data storage and processing. Distributed databases are deployed on self‐organizing network servers or dispersed across the Internet, enterprise networks, Internet, and other independent computers in self‐organizing networks. Data is stored on multiple machines, and operations are not limited to a single machine but allow transactions to be executed across multiple machines to improve database access performance.
Distributed databases have become a core technology for big data processing. Based on their structure, distributed databases include homogeneous and heterogeneous systems. The former has database instances running in environments with the same software and hardware, featuring a single‐access interface. The latter operates in environments where hardware, operating systems, database management systems, and data models vary. Based on the types of data processed, distributed databases mainly include relational (such as Structured Query Language, SQL), nonrelational (such as NoSQL), extensible markup language (XML)‐based, and NewSQL distributed databases. Among these, NoSQL and NewSQL distributed databases are the most widely used [16]. NoSQL distributed databases, designed to meet the demands for high concurrency, efficient storage access, high reliability, and scalability in big data environments, are divided into key‐value stores, column stores, document‐oriented databases, and graph databases. NewSQL distributed databases, characterized by real‐time processing, complex analysis, and fast querying, are relational distributed databases designed for massive data storage in big data environments, including Google Spanner, Clustrix, and VoltDB. SQL‐distributed databases are relational distributed databases, with typical examples including Microsoft's and Oracle's distributed databases. XML‐based distributed databases mainly store data in XML format and are essentially document‐oriented, similar to NoSQL distributed databases [22].
Compared to edge computing models, distributed databases provide data storage in big data environments but pay less attention to the heterogeneous computing and storage capabilities of the devices they reside on, focusing mainly on achieving distributed data storage and sharing. Distributed database technology requires significant space and offers lower data privacy. For distributed transaction processing across multiple databases, data consistency technology is a major challenge for distributed databases [14]. In edge computing models, data resides on edge devices, offering higher privacy, reliability, and availability. In the era of the IoE, “heterogeneous edge architectures and the need to support multiple application services” will become the fundamental approach for edge computing models to handle big data processing.
1.2.1.2 Peer‐to‐Peer (P2P) Computing Models
P2P computing [27] is one of the early file transfer technologies that pushed computing to the edge of the network. The term P2P was first introduced in 2000 to implement file‐sharing systems. Since then, it has gradually developed into an important subfield of distributed systems. The key research topics in P2P models include decentralization, maximizing scalability, tolerance of high‐level node churn, and preventing malicious behavior. Major achievements in this field include:
Distributed Hash Table (DHT), which later evolved into the general paradigm for key‐value distributed storage in cloud computing models.
Generalized gossip protocols, which have been widely used for complex task processing applications beyond simple information dissemination, such as data fusion and topology management.
Multimedia streaming technology, in forms such as video on demand, real‐time video, and personal communication.
However, widespread media coverage of P2P being used for illegal file sharing and related lawsuits has hindered the practical recognition of some commercial technologies based on the P2P model.
The edge computing model bears significant similarities to P2P technology, while it expands on the latter with new technologies and methods, extending the concept of P2P to network edge devices. This represents a fusion of P2P computing and cloud computing.
1.2.1.3 Content Delivery Network (CDN) Models
CDN [29] was proposed by Akamai in 1998. CDN is an Internet‐based caching network, which relies on caching servers deployed in different places and points users' access to the nearest caching server through load balancing, content distribution, scheduling, and other functional modules of the central platform. Therefore, CDN can reduce network congestion and improve user access response speed and hit rate. It has gained significant attention from both academia and industry since it was proposed. Companies like Amazon [2] and Akamai [1] possess mature CDN technologies that provide users with the expected performance and experience while reducing the operational pressures on service providers.
Active content distribution networks (ACDNs), an improvement over traditional CDNs, help content providers avoid the hassle of predicting the preconfiguring resources and determining their locations [30]. ACDN allows applications to be deployed on any server and uses newly designed algorithms to replicate and migrate applications between servers as needed.
The concept of edge computing can be traced back to around the year 2000, when CDNs were deployed large scale. At that time, major companies like Akamai announced the distribution of web‐based content through CDN edge servers. The primary goal of this method was to benefit from the short distances and available resources of CDNs to achieve large‐scale scalability. In the early days of edge computing, the “edge” was limited to CDN cache servers distributed around the world. However, today's development of edge computing has far exceeded the scope of CDNs. The “edge” in the edge computing model is not confined to edge nodes; it includes any computational, storage, and networking resources along the path from data sources to cloud computing centers.
1.2.1.4 Function Cache and Cloudlet
To enable static content distribution, CDN emphasizes the backup and caching of data, while edge computing focuses more on function caching to improve computational capabilities. Function cache was proposed by Ravi et al. [31], where it is applied to personalized mailbox management services to save latency and bandwidth. Satyanarayanan et al. [33] introduced the concept of Cloudlet, which is a trusted, small‐scale, and resource‐rich host, located at the edge of the network, connected to the Internet, and can be accessed by mobile devices to provide services. Cloudlet is also known as “small cloud” as it can provide services for users, similar to the cloud server. At this point, edge computing focused on the downstream transfer of functions from cloud servers to edge servers, aiming to reduce bandwidth usage and minimize delays.
1.2.1.5 Mobile Edge Computing
The development of the IoE has enabled the interconnection of numerous types of devices, such as smartphones, tablets, wireless sensors, and wearable devices. However, the limited energy and computing resources of most network edge devices make the design of IoE particularly challenging. Mobile edge computing (MEC) [19] is a new network architecture that provides information technology services and cloud computing capabilities within the proximity of the mobile user's wireless access network. It has become a standardized and regulated technology. In 2014, the European Telecommunications Standards Institute (ETSI) introduced the standardization of the term MEC, highlighting that MEC provides a new ecosystem and value chain. Utilizing MEC, intensive mobile computing tasks can be offloaded to nearby network edge servers. Because MEC is located within the wireless access network and close to mobile users, it can achieve lower latency and higher bandwidth, thereby improving service quality and user experience. MEC is also a key technology in the development of 5G, helping to meet the high standards of 5G in terms of latency, programmability, and scalability. By deploying services and caches at the network edge, MEC reduces congestion in the core network and efficiently responds to use requests.
Task migration is one of the challenges in mobile computing technology, particularly in environments where continuous service availability and seamless user experience are crucial. The process involves transferring ongoing tasks from one computational node to another, which can be triggered by various factors such as device mobility, energy conservation needs, or load balancing requirements. Effective task migration must minimize latency, avoid data loss, and maintain application state continuity, which is challenging due to the heterogeneous and dynamic nature of mobile environments. Furthermore, ensuring security during data transfer, managing the energy consumption of mobile devices, and dealing with fluctuating network conditions are additional hurdles that need optimization solutions. As mobile computing continues to evolve, developing robust, efficient, and secure task migration mechanisms will be critical to fully leveraging the potential of mobile platforms. MEC has been applied in various scenarios, such as vehicular networks, IoT gateways, auxiliary computing, intelligent video acceleration, and mobile big data analysis.
MEC emphasized the establishment of edge servers between the cloud server and edge devices to process computing. However, mobile edge nodes are generally considered to lack computing capabilities. In contrast, the nodes in the edge computing model possess strong computing capabilities. Therefore, MEC resembles the architecture and hierarchy of an edge computing server, functioning as an important part of edge computing.
1.2.1.6 Fog Computing
Cisco introduced fog computing in 2012 and defined fog computing as a highly virtualized computing platform for migrating cloud computing center tasks to network edge devices [7]. Fog computing provides computing, storage, and network services between end devices and traditional cloud computing centers, complementing cloud computing. Vaquero and Rodero‐Merino [39] have provided a comprehensive definition of fog computing, which extends cloud‐based network architecture by introducing an intermediate layer between the cloud and mobile devices. This intermediate layer, known as the fog layer, consists of fog servers deployed at the network edge. Fog computing reduces the need for multiple communications between the cloud computing center and mobile users. It relieves the bandwidth load and energy consumption pressure of main links by reducing the number of communications between cloud computing centers and mobile users. When there is a large volume of mobile users, they can access cached content and request specific services from the fog computing servers. Additionally, fog computing servers can interconnect with cloud computing centers, leveraging their powerful computational capabilities and extensive applications and services.
The concepts of edge computing and fog computing have great similarities and often represent the same idea. If we are to distinguish between the two, this book posits that edge computing, in addition to focusing on infrastructure, also pays attention to edge devices and places more emphasis on the design and implementation of edge intelligence. In contrast, fog computing focuses more on the management of back‐end distributed shared resources. As shown in Figure 1.4, since 2017, the level of attention to edge computing has gradually surpassed that of fog computing, and its attention continues to rise.
Figure 1.4 “Edge computing” and “fog computing” trends.
1.2.1.7 Cloud‐Sea Computing
In the context of IoE, the amount of data to be processed will reach ZB levels. The sensing, transmission, storage, and processing capabilities of information systems need to be correspondingly enhanced. To address this challenge, in 2012, the Chinese Academy of Sciences launched a ten‐year strategic priority research initiative called the Next Generation Information and Communication Technology (NICT) initiative. Its main purpose is to carry out research on the “Cloud‐Sea Computing System Project” [40]. It aims to augment cloud computing by cooperation and integration of the “cloud computing” system and the “sea computing” system. “Sea” refers to an augmented client‐side consisting of human‐facing and physical world‐facing devices and subsystems. The research focuses on proposing system‐level solutions from perspectives such as overall system architecture, data center and server and storage system layers, and processor chip level.
Cloud‐sea computing focuses on the two ends “sea” and “cloud” while edge computing focuses on the data path between “sea” and “cloud.” Cloud‐sea computing is a great subset example of edge computing.
1.2.2 Rapid Growth Period
Since 2015, edge commuting has been in a rapid growth period, attracting intensive close attention from academia and industry.
At the government level, in May 2016, the National Science Foundation (NSF) listed edge computing as one of the highlighted areas in the research of computer systems. In August 2016, NSF and Intel formed a partnership in information center networks in wireless edge networks (ICN‐WEN) [37]. In October 2016, the NSF held the NSF Workshop on Grand Challenges in edge computing [8]. The workshop focused on three topics: the vision of edge computing in the next five to ten years, the grand challenges to achieving the vision, and the best mechanisms for academia, industry, and the government to attack these challenges in a cooperative way. This indicates that the development of edge computing has attracted great attention at the government level.
In academia, a formal definition of edge computing is given in the paper Edge computing: vision and challenges[34]. Edge computing is defined as enabling technologies that allow computation to be performed at the edge of the network, processing downstream data on behalf of cloud services, and upstream data on behalf of IoT services. This paper pointed out the challenges of edge computing and is one of the most cited papers in the edge computing field. In October 2016, ACM and IEEE jointly organized the first ACM/IEEE Symposium on Edge Computing (SEC). Since then, the International Conference on Distributed Computing Systems (ICDCS), the International Conference on Computer Communications (INFOCOM), the International Middleware Conference, and other important international conferences have added an edge computing track and/or workshops to their main conferences.
At the same time, multiple industry sectors have actively promoted the development of edge computing. In September 2015, ETSI published a white paper on MEC. In November 2015, Cisco, ARM, Dell, Intel, Microsoft, and Princeton University jointly established the OpenFog Consortium, which is dedicated to the development of Fog Reference Architecture [28]. The OpenFog Consortium merged into the Industrial Internet‐of‐Things (IIoT) in January 2019. In November 2016, Huawei, Shenyang Institute of Automation of Chinese Academy of Sciences, China Academy of Information and Communications Technology (CAICT), Intel, ARM, and iSoftStone established the Edge Computing Consortium (ECC) in Beijing, China, which is dedicated to advancing cooperation among industry resources from government, vendor, academic, research, and customer sectors, and pushing forward the sustainable development of the edge computing industry [12]. In March 2017, the ETSI MEC Industry Specification Working Group was formally renamed to multiaccess edge computing, aiming to better meet the requirements of edge computing and related standards. Linux EdgeX Foundry was also built in 2017; it is a vendor‐neutral open‐source project hosted by The Linux Foundation. It aims to build a common open framework for IoT edge computing. In January 2018, Automotive ECC (AECC) was established to drive the network and computing infrastructure needs of automotive big data [4], which indicates that edge computing is valued in the vehicle domain. In the same year, the Cloud Native Computing Foundation (CNCF) Foundation and Eclipse Foundation cooperated to bring Kubernetes, which has been widely used in the ultra large‐scale cloud computing environment, into the edge computing scene of the IoT. Subsequently, KubeEdge, a Kubernetes native edge computing framework, was accepted into the CNCF sandbox in March 2019 [24]. In April 2019, the Bio‐IT World Conference and Expos added the edge track [6], which means that edge computing is important to the health domain as well.
In the rapid growth period, the industry has seen significant advancements, evidenced by the availability of multiple edge environment solutions from major service providers. Today, options like AWS Greengrass [5], Microsoft Azure [25, 26], Google Cloud Platform Edge Zones [15] have made it easier for businesses and developers to deploy and manage edge computing infrastructures effectively. These developments underscore the maturity and widespread adoption of edge computing across various sectors.
1.2.3 Intelligence Integration Period
Edge computing has seen substantial growth and transformation in recent years, driven by the increasing demand for low‐latency data processing and efficient resource utilization. The development of edge computing is characterized by bringing together IoT, big data, and mobile computing into an integrated and ubiquitous computing platform. The capability of delivering on‐demand computing power at the edge and processing a vast amount of data from various devices/sensors enables real‐time analytics and decision‐making. A significant advancement within this domain is the integration of edge intelligence, where artificial intelligence (AI) and machine learning (ML) algorithms are deployed at the edge. This symbiotic relationship enhances the capability of edge computing, allowing for sophisticated data analysis and autonomous decision‐making directly at the data source. Edge intelligence empowers devices to process and act on data locally, leading to smarter, faster, and more efficient systems across various industries, from autonomous vehicles to smart cities and beyond.
Building on this foundation, the intelligence integration period marks a crucial phase in the evolution of edge computing. Technological strategies such as pruning, quantization, and knowledge distillation are employed to optimize AI models for efficient operation on edge devices. Simultaneously, AI algorithms find wide application across systems such as smart surveillance, autonomous vehicles, health monitoring systems, industrial IoT, smart agriculture, and retail enhancements, further demonstrating the pervasive impact of this integration. These advancements not only improve responsiveness but also deliver substantial societal benefits. While the potential of these integrations is immense, the associated privacy and security concerns are non‐negligible and will be discussed in Chapter 8, with deeper technological and application‐based discussions slated for Chapters 3, 4, and 7.
1.3 What Is Edge Computing?
After exploring the background of edge computing's emergence and examining its three distinct phases of development, it's time to address a fundamental question: what exactly is edge computing?
There is no standard definition for edge computing yet. In the field of edge computing, industry experts and other researchers have provided their own definitions. For example, IBM views edge computing as a distributed computing framework that brings enterprise applications closer to data sources such as IoT devices or local edge servers [20]. CISCO interprets edge computing as a model that shifts computing resources from central data centers or public clouds closer to devices, that is, embedded at the edge of service provider networks [9]. Satyanarayanan defines edge computing as a computing paradigm in which substantial computing and storage resources—variously referred to as cloudlets, micro data centers, or fog nodes—are placed at the Internet's edge in close proximity to mobile devices or sensors [32]. Yousefpour et al. believe edge computing is located at the edge of the network close to IoT devices, and edge can be more than one hop away from IoT devices in the local IoT network [41].
Figure 1.5 Edge computing paradigm.
Here, we give our own definition of edge computing. In the vision paper published in 2016, we highlighted that edge computing refers to the enabling technologies allowing computation to be performed at the edge of the network, on downstream data on behalf of cloud services and upstream data on behalf of IoT services.[34]. As shown in Figure 1.5, in our definition, “edge” can be any computing and network resources along the path between data sources and cloud data centers. The rationale of edge computing is that computing should happen at the proximity of data source, close to the users. We can interpret the word “close” in two ways. First, the edge computing resource and the end users may be close in the communication network. Therefore, the small network size makes it more feasible to deal with network instability (e.g., bandwidth, delay, and jitter). Second, the resource and users may be close in spatial distance, which means they share similar environmental information. The computing resources may leverage the shared information to provide personalized services and improve the user experience. Network distance and spatial distance are not correlated to each other, and it may depend on the concrete scenarios to decide which type of close (or both) is appropriate.
If we view resources on the path between IoT services and cloud services as a continuum, edge can be any computing, storage, and network resources on this path. Depending on the specific requirements and concrete scenarios, the edge can be one or multiple resources (nodes), as shown in Figure 1.6. It can be a smartphone or a desktop serving as the edge between body things and the cloud, a gateway in a smart home as the edge between home things and the cloud. It can also be as small as embedded devices such as wearable sensors and security cameras, or as big as a micro data center. There are a huge amount of edge resources; they are scarcely distributed around end users, independent of each other. Edge computing is dedicated to unifying these resources that are close to end users in either the communication network or spatial distance and provides computing, storage, and network services for applications.
Figure 1.6 Edge computing is a continuum.
If we understand edge computing from a biological perspective, we can make the following analogy: cloud computing is akin to the human brain, while edge computing is akin to nerve endings. When a needle pricks the hand, a person instinctively pulls their hand back before the brain even realizes the prick because the reflex action is processed by the nerve endings. This reflex action speeds up the response, preventing further harm, while allowing the brain to focus on more complex tasks. In the future era of the IoE, it is impractical for cloud computing to act as the “brain” for every device. Instead, edge computing allows edge devices to have their own “brains.”
To have an overview of the development of edge computing, we investigate the time period when it got attention and became popular. If you search for “edge computing” in Google Trends, set the time range to be 01/01/2010–now, you will see a trend like Figure 1.7. “Interest over time” shows search interest relative to the highest point on the chart for the given region (in this example, we choose Worldwide) and time (from 2010 till 2024). A value of 100 means the term is in the peak popularity. A value of 50 means that the term is half as popular. A score of 0 means there was not enough data for this term. We can see that the attention to edge computing has been continuously increasing since 2016, reflecting its importance in the development of technology. In the years 2022 to 2023, there is a drop in “interest over time” for edge computing. At the same, another term, Edge AI, has been raising attention and has been getting more and more interest since 2023, when the large language model (LLM) started to dominate the artificial intelligence (AI) and machine learning (ML) market. The trend of these two words perfectly shows the focus of the research area in edge computing. Before LLM showed up, the research focus in this field was on the computing paradigm itself, such as the architectures and components of edge computing (Chapter 3), edge computing hardware and software (Sections 4.1 and 4.2), and challenges and solutions in edge computing (Chapter 5). With the development of AI and ML, especially LLMs, the world has witnessed the power of AI models. However, to make these powerful models accessible to the masses, the computing and storage resource constraints became a significant bottleneck. Therefore, the research focus in this field has shifted to enabling edge‐based AI by tackling the problem of resource constraints (Sections 4.3 and 4.4).
Figure 1.7 “Edge computing” and “edge AI” trends.
1.4 Summary and Practice
1.4.1 Summary
This chapter provides the definition and core concept of edge computing, emphasizing that edge computing is a continuum. The “edge” in edge computing refers to any computing, storage, and network resources along the path from the data source to the cloud computing center. The discussion of the development and challenges of big data processing and the Internet of Everything helps to understand the background of the emergence of this computing paradigm. This chapter also reviews the historical development of data‐oriented computing models such as distributed databases, P2P, CDN, MEC, fog computing, and Cloud‐Sea computing. Additionally, it introduces the current status of edge computing and its close connection with edge intelligence.
1.4.2 Practice Questions
What is the “edge”?
What are the main characteristics that distinguish edge computing from traditional cloud computing?
Identify and explain the challenges that traditional cloud computing faces in handling the large volumes of data generated by IoE devices.
Why is edge computing necessary in the era of the Internet of Everything?
Explain examples of real‐world applications for each use case presented in
Figure 1.6
.
Based on the background and evolutionary history of edge computing discussed in this chapter, what do you think are the key challenges that could arise during the development of edge computing?
1.4.3 Course Projects
Analyze real‐world case studies where edge computing is used to solve specific problems and understand why edge computing was necessary in each case. Case studies can be found from sources like
https://lfedge.org/
.
Conduct a comprehensive study of various open‐source edge platforms to understand the capabilities, strengths, weaknesses, and potential use cases of each platform. The platforms to be researched could include, but are not limited to: LF Edge Projects (EdgeX Foundry, Akraino, and Open Horizon), Kubernetes for edge (KubeEdge, K3s, and MicroK8s), OpenFaaS.
