Edge Computing: Models, technologies and applications
2: College of Computer Science and Technology, Zhejiang University, Hangzhou, China
Cloud computing has evolved as a cost-effective, easy-to-use, elastic and scalable computing paradigm to transform today's business models. 5G, Industrial IoT, Industry 4.0 and China-2050 frameworks and technologies are introducing new challenges that cannot be solved efficiently using current cloud architectures. To handle the collected information from such a vast number of devices and actuators, and address these issues, novel concepts have been proposed to bring cloud-like resources closer to end users at the edge of the network, a technology called edge computing. From architectures to models, technologies and applications, this book focuses on the Edge Computing paradigm due to its unique characteristics where heterogeneous devices can be equipped with decision making processes and automation procedures to carry out applications across widely geographically distributed areas. This book provides valuable insights for ICTs engineers, scientists, researchers, developers and practitioners who are involved in developing and implementing edge and cloud-based solutions ranging from sensors and actuators to cloud-based back-end systems.
Inspec keywords: cloud computing; distributed processing
Other keywords: virtualization; decision-making processes; orchestration; edge computing; networking; theoretical models; heterogeneous devices; smart cities; widely geographically distributed areas; cloud computing
Subjects: Computer networks and techniques; Distributed systems software; General and management topics
- Book DOI: 10.1049/PBPC033E
- Chapter DOI: 10.1049/PBPC033E
- ISBN: 9781785619403
- e-ISBN: 9781785619410
- Page count: 477
- Format: PDF
-
Front Matter
- + Show details - Hide details
-
p.
(1)
-
Part I. Models
1 Introduction to edge computing
- + Show details - Hide details
-
p.
3
–25
(23)
Edge computing is the model that extends cloud computing services to the edge of the network. This model aims to move decision-making operations as close as possible to data sources since it acts as an intermediate layer connecting cloud data centres to edge devices/sensors. Transferring all the data from the network edge to the cloud data centres for processing may create a latency problem and outstrip the network's bandwidth capacity. To resolve this issue, it might be best to process data closer to the devices/sensors. This chapter will take a deep dive into edge computing, its applications, and the existing challenges related to this model.
2 Edge computing architectures
- + Show details - Hide details
-
p.
27
–44
(18)
By considering an edge computing platform as one of the main computing tiers in a fully comprehensive cloudisation model, this chapter first discusses the distinguished features of edge computing compared to other platforms from the perspectives of service, operation and control. In addition, a local view of the intrinsic architecture of an edge computing host is analysed. Subsequently, a standard reference architecture of the edge computing platform is presented, which defines the internal and external interfaces among components at the computing host and system levels. The reference architecture is matched to the network function virtualisation management and orchestration (NFV MANO) model for consideration as a virtualised network function (VNF). In particular, the position and roles of VNF edge computing in the standard mobile network reference model are demonstrated. Finally, the open issues and challenges are presented by way of conclusion.
3 Big data analytical models for/on edge computing
- + Show details - Hide details
-
p.
45
–52
(8)
In this chapter, the author presented a holistic approach to the concept of big data modeling towards edge computing. In the first part of the analysis, the author introduced a detailed introduction, critical challenges in the field of big data analytics, and how those challenges can impact edge applications deployed on the cloud. In the third section, the author presented the technical aspects of in-memory databases that can speed up the data computation for the low-power edge devices. In the fourth part, the author highlighted a few of the security challenges and mechanisms to overcome the problems during the system design phase. Nevertheless, many open issues and challenges can arrive in the field of edge computing as it's an advanced stream that has started to flourish across the various domains
4 Data security and privacy models for/on edge computing
- + Show details - Hide details
-
p.
53
–75
(23)
As a new computing paradigm, edge computing (EC) has gained much popularity in various delay-sensitive application scenarios (like vehicular networks and autonomous driving). With the prevalence of EC, its security and privacy issues become more and more significant for its further development. When referring to the security and privacy issues in EC, its meaning is twofold: the first is the inherent security issues of EC architectures and the other is how we can strengthen the security of current systems by introducing the paradigm of EC. Thus in this chapter, after analyzing the unique properties of EC, intrusion, lightweight authentication, access control and private information leakage are chosen to be four typical threats and challenges for the security or privacy issues of EC. Then, some practical security and privacy solutions prompted by EC in existing research works are present. After that, several open research issues are given to draw more attentions from academia and industry to guarantee the security of this new paradigm.
5 Networking models and protocols for/on edge computing
- + Show details - Hide details
-
p.
77
–95
(19)
Edge computing is a distributed computing paradigm which brings the computing infrastructure at the edge of the network and close to the users in order to reduce the latency and improve the user experience. It is the realisation of a concept of an intermediate layer between the users and cloud computing infrastructure to accommodate explosive growth of data generated by underlying user devices and to process it with lesser response time than cloud computing systems. Mobile network operators are the potential providers of edge computing as a service by using the infrastructure of their service networks existing in close proximity to users. In order to establish the synergy between the devices or resources of user end, edge layer and cloud layer, a thorough understanding of networking models and communication protocols is required. This will help the researchers and system designers to design optimised and efficient edge computing systems, which will improve the user experience with minimum incurred costs. This chapter provides an overview of networking/reference models and the corresponding communication protocols for edge computing systems. First, the chapter starts with a layered system architecture representing interaction between the user devices with cloud computing infrastructure via fog layer followed by reference models corresponding to each layer. Description of communication protocols corresponding to each reference model is also provided besides a taxonomy classifying all the discussed communication protocols on the basis of their characteristics.
6 Computing and storage models for edge computing
- + Show details - Hide details
-
p.
97
–123
(27)
In this chapter, the author have examined the value proposition of such an approach, and the complementary benefits of edge, fog, and cloud resources. The author have explored different computing and storage models to support application composition and data storage on edge resources, alongside fog and cloud. There are also existing techniques and runtime strategies from cloud computing, Big Data platforms, distributed storage systems, and P2P systems that can be suitably extended and adapted to meet these unique needs. The author have offered two case studies of an edge-based video analytics platform for tracking, and an edge-centric storage system, which highlight these design elements. Lastly, the author have highlighted open problems in the context of computing and storage models for the edge.
7 Resource allocation models in/for edge computing
- + Show details - Hide details
-
p.
125
–146
(22)
During last decades, cloud computing enabled centralization of computing, storage and network management through cloud data centres, cellular core networks and backbone IP networks. Unlimited computing resources and storage can be used to provide elastic cloud services to a vast range of customers from organizations encompassing hundreds of servers to resource-constrained end users. Despite numerous advantages, Cloud computing is facing increasing limitations such as remarkable communication delay, jitter and network traffic caused due to far located data centres from end users. Edge computing has emerged in order to shift the traditional cloud model towards a decentralized paradigm to deal with strict requirements of latency, mobility and localization of new systems and applications (e.g. Internet of Things (IoT)). Edge computing places the content and resources near to end users to ensure a better user experience. Edge computing is still at the stage of development and encounters many challenges such as network architecture design, fault-tolerance and distributed service management, cloud-edge interoperability and resource allocation. In this chapter, we present an overview on existing resource allocation models proposed for and leveraged in edge computing. Specifically, we explore the key concepts of resource allocation in edge computing, and investigate recent researches have been conducted in this domain along with their optimization models. Finally, we discuss challenges and open issues.
8 Human-in-the-loop models for multi-access edge computing
- + Show details - Hide details
-
p.
147
–168
(22)
In this chapter, studied the role of HITL-centric networks based on the concept of MEC-enabled FiWi-enhanced LTE-A HetNets in realizing the Tactile Internet's <10 -ms delay challenge, thereby paying particular attention to coexistent local/nonlocal teleoperation and inquiring into the specific characteristics of haptic traces at the packet level. The Tactile Internet traffic analysis reveals that the teleoperation command and feedback paths can be jointly modeled by the GP, gamma, or deterministic packet inter-arrival time distribution, depending on the value of DC parameter d. DC was shown to be particularly effective in the command path. Alternatively, in the feedback path, the proposed AI-based sample forecasting module embedded in MEC servers is instrumental in achieving a very high haptic sample forecasting accuracy with an MSE of zero in case of local teleoperation.
-
Part II. Technologies
9 Distributed big data computing platforms for edge computing
- + Show details - Hide details
-
p.
171
–195
(25)
In this chapter, we present and analyze some popular platforms for Big Data processing, with a focus on edge computing. We start with a review of edge computing, followed by a classification of frameworks and models for Big Data processing. In the next section, we describe some well-known platforms for Big Data processing, such as MapReduce, Spark, Flink and Google Cloud Dataflow with its open-source version under Apache Beam. In different section, we present Big Data frameworks specific to edge computing, including hybrid MapReduce and Apache Edgent. Finally, we summarize the chapter by presenting challenges and opportunities for research in this area.
10 Distributed execution platforms for edge computing
- + Show details - Hide details
-
p.
197
–222
(26)
In this chapter, the authors present an edge computing framework to harvest and orchestrate distributed edge resources for executing complex applications, e.g., video analytics, that overwhelms the back end, while infeasible on a single smart device. This aims at maximizing the gain from edge data and enabling a new set of applications. For example, recruit a set of microphones and CPUs on restaurant patron phones in Manhattan, NY, and re-purpose them to find the least noisy or crowded restaurant. In this scenario, and many others, edge computing enables decisions based on live data, unlike static current offerings based on address, pictures, and customer reviews.
11 Collaborative platforms and technologies for edge computing
- + Show details - Hide details
-
p.
223
–247
(25)
This chapter focuses on collaborative edge computing, and tries to figure out how collaborations are achieved in edge computing from the following aspects: (1) edge computing: ecosystem and players; (2) computing and networking collaborations in edge computing; (3) use cases and applications in collaborative edge computing; and (4) platforms and prototypes of collaborative edge computing.
12 Serverless architecture for edge computing
- + Show details - Hide details
-
p.
249
–264
(16)
In this chapter, present a review over two emerging technologies, namely edge computing and serverless computing. Both technologies have great potential to be extended and adopted in next few years for development of smart applications. In addition, there are many cloud providers working toward enabling users to get benefit from these services. We strongly believe that combining these two technologies provides a number of benefits for developers to be able to use the edge -cloud computing systems with less complexity and effort. This paper presented a conceptual framework along with a case study to show how the serverless edge computing architecture can help to develop smart applications. There are several technical challenges that should be investigated in this area, which we have presented as the open questions for future work.
13 Open-source projects for edge computing
- + Show details - Hide details
-
p.
265
–290
(26)
In this chapter, the author covered an overview to the EC technologies as well as a scope classification to its entire paradigm. The author also reviewed the state-of-the-art reference architectures and standardization, as well as top ten open-source projects and platforms in EC. Moreover, the author mentioned open issues and challenges in the EC paradigm and discussed them in detail.
14 Simulators and emulators for edge computing
- + Show details - Hide details
-
p.
291
–311
(21)
In this chapter, we perform a study of the existing tools for the evaluation of Fog/Edge infrastructures. First, we analyze the state of the art in the simulation of Fog/Edge infrastructures and determine the main challenges in simulation and modeling such infrastructures. Then, we use a scientific methodology to identify the most important simulation and emulation tools, identifying their main characteristics, and define a classification. Each tool is then described in detail, and compared with the others. Finally, we conclude the chapter with a discussion about future research directions in the area.
-
Part III. Applications
15 Smart cities enabled by edge computing
- + Show details - Hide details
-
p.
315
–336
(22)
The development of smart cities is inseparable from the application of edge computing technology. The value of edge computing in smart cities is reflected in specific application scenarios. So, this chapter mainly introduces the definition of smart cities and their architecture, and enumerates some applications with edge computing technology. It also introduces how edge computing technology is applied in dealing with urban traffic congestion. Next, the computation offloading, resource allocation and task scheduling problems in edge computing -enabled smart city are discussed according to the academic research. Finally, we discuss the security and privacy problem in edge computing -based smart city. The application of edge computing in smart cities is much more than that mentioned in this chapter. With experts and scholars in different domain investing in edge computing research, the role of edge computing in smart cities will be further enhanced.
16 Smart healthcare systems enabled by edge computing
- + Show details - Hide details
-
p.
337
–356
(20)
Proliferation of mIoT devices and mobile communication technology along with intelligence services prompt the emergence of edge -based smart healthcare system. This irmovation results in the reformation of the healthcare system from hospital oriented to patient oriented. Edge computing makes modern healthcare system more time critical and sensitive through reducing the time waste in data transmission compared to the stand-alone cloud -based system. And coupled with the intelligence services, it can help doctors to make smart decision during emergency cases. In this chapter, we first look back to the evolution process of modern healthcare systems and give a detailed introduction about the mloT. After that, a three -layer modern healthcare system enabled by edge computing was proposed in this paper, and it is a good approach to deploy and implement patient -oriented distributed healthcare system to improve precision medicine services. And then, we present some successful applications of daily health monitor system in the market. Finally, we highlight several open issues and challenges associated with the implementation of the edge -based healthcare system. In future, we can integrate this smart healthcare system into the application of smart city, and figure out how to cooperate with other parts in smart city, such as smart transportation and smart building to make our life more convenient.
17 Smart hospitals enabled by edge computing
- + Show details - Hide details
-
p.
357
–379
(23)
In last few years, achievements in information and communications technologies (ICTs), such as the electronic health record (EHR), have improved the healthcare system. However to be effective, paramedics and doctors have to consult the most recent version of EHRs anytime and anywhere. A possible solution is to store EHRs on remote storage services. However, the EU General Data Protection Regulation (GDPR) does not allow to store plain files containing personal data in services accessible remotely. To solve this challenge, a possible solution is to use Edge computing devices running Secret Sharing algorithms to split and merge EHRs on demand; however, these techniques have not been evaluated before for these purposes. To address this issue, in this work we analyse the redundant residue number system (RRNS). In particular, considering different EHR sizes (from 10kB to 1 MB), we evaluated computation time (split and recomposition), transfer time (upload and download) from/to public Cloud storage providers (Google Drive, Mega and Dropbox) and storage requirement. Results showed that, in configuration with seven levels of redundancy, the RRNS uses only 50% of the storage required for the simple file replication. We also discovered that Google Drive, due to synchronization overhead, is slower than other Cloud service providers for the upload of chunks but faster for the download.
18 Smart grids enabled by edge computing
- + Show details - Hide details
-
p.
381
–408
(28)
A smart grid is the nervous system of the power generation, transmission, and distribution systems that makes a great use of the information and communications technologies (ICTs). The ICT enables the smart grid to timely detect, monitor, and react to local changes in usage and in the event of electrical faults of various types. The smart grid is the nexus of distributed electrical sensors, smart energy meters, smart appliances often deployed in customer premises, transducers, network interfaces, remote terminals, servers, and a multiplexed communication system which transmits data and commands between parts installed across the entire power grid system components. The main power grid components include the power generation station, high-voltage transmission system, distribution systems, and customer premises. The sensors that can be interconnected to one another using various network architectures and computing paradigms are the eyes and ears of the smart grid which provide information vital for efficient and timely fault detection, monitoring, and controlling the entire power grid system. Hence, the smart grid is derived from the general-purpose network architecture and computing models in a manner as to fit the purposes of the electrical grid system. The main thing that distinguishes a smart grid from the general-purpose computer network is that it is one specific application of it. The most striking characteristic of computer networks is their generality. They are not optimized for a specific application like the smart grid. They are built principally from general-purpose programmable hardware capable of carrying and supporting many different types of data, and a wide spectrum of ever growing applications. Just like the general-purpose computer network, the smart grid could be deployed in client-server, peer-to-peer, or distributed architecture. In a similar fashion, the computational paradigm of the smart grid could be cloud, fog, or edge based. But the smart grid is typically the embodiment of the Internet of Things (IoT) or cyber-physical systems; hence, the most suitable computing paradigm is one that brings the computation and data storage closer to the point where data are created and garnered. Thus, this chapter looks at the typical ways how the edge computing paradigm is applied to improve reliability, the load forecasting capability, security and privacy of the smart grid. To put it another way, this chapter focuses on four things. It, first, lays down the foundations and background knowledge about the power grid, smart grid, and edge computing paradigm. Second, it explains the factors that affect the reliability of the smart grid and explains the ways how the edge computing techniques can improve the reliability of the smart grid. Third, it explores the requirements and ways how power consumption prediction could be accurately performed at the edge using artificial intelligence (AI), machine learning (ML), and deep learning (DL) methods coupled with advanced electrical signal processing techniques. Finally, it presents how the security and privacy issues of a smart grid enabled by edge computing could be addressed.
19 Smart surveillance for public safety enabled by edge computing
- + Show details - Hide details
-
p.
409
–433
(25)
To address issues in traditional security solutions relying on centralized third party authority, a blockchain-enabled decentralized security services framework is proposed, and it provides a secure data sharing and AC mechanism for heterogeneous, distributed smart surveillance system. Leveraging the fine -granularity and loose -coupling features of the microservices architecture, the security functionalities are decoupled into multiple containerized microservices that are computationally affordable to each individual device. The brief experimental results validated that the blockchain-enabled microservices solution is able to efficiently and effectively enforce security in a distributed IoT system.
-
Back Matter
- + Show details - Hide details
-
p.
(1)