Intelligent Edge Computing for Cyber Physical Applications

By Mohamed Elhoseny

Intelligent Edge Computing for Cyber Physical Applications introduces state-of-the-art research methodologies, tools and techniques, challenges, and solutions with further research opportunities in the area of edge-based cyber-physical systems. The book presents a comprehensive review of recent literature and analysis of different techniques for building edge-based CPS. In addition, it describes how edge-based CPS can be built to seamlessly interact with physical machines for optimal performance, covering various aspects of edge computing architectures for dynamic resource provisioning, mobile edge computing, energy saving scenarios, and different security issues.

Sections feature practical use cases of edge-computing which will help readers understand the workings of edge-based systems in detail, taking into account the need to present intellectual challenges while appealing to a broad readership, including academic researchers, practicing engineers and managers, and graduate students.

• Introduces and provides reviews on cyber physical and edge computing systems, with different architectures and models needed to address sustainable solutions to social, environmental and economic applications
• Presents the different architectures of edge computing for building cyber physical systems with dynamic resource provisioning and security solutions
• Provides AI based perspectives to edge-based cyber physical systems with different algorithms and AI based security solutions
• Covers different case studies and applications in detail, with real-life examples and possible challenges that can be encountered
• Offers perspectives for the design, development and commissioning of intelligent edge-based cyber physical systems

• Language: English
• Publisher: Academic Press
• Release date: Feb 3, 2023
• ISBN: 9780323994330

Book preview

Chapter 1

Introduction to different computing paradigms: cloud computing, fog computing, and edge computing

Swati Vijay Shinde¹, D. Jude Hemanth² and Mohamed Elhoseny³,    ¹Pimpri Chinchwad College of Engineering, Pune, Maharashtra, India,    ²Department of ECE, Karunya Institute of Technology, Chennai, Tamil Nadu, India,    ³College of Computer Information Technology, American University in the Emirates, Dubai, UAE

Abstract

This chapter introduces the different computing paradigms: Cloud computing, fog computing, and edge computing, with emphasis on edge computing. Each of these paradigms is more suited to a particular application than the other, and each has its own features. The different computing paradigms are deployed depending upon the need of the application. Cloud computing systems have proved to be useful in almost all application areas, providing remote processing and storage of the end-user data and software in the bigger data centers. However, for smaller requests, accessing a remote data center has resulted in latency and bandwidth problems, which become particularly evident in IoT-based applications, where sensor readings are uploaded to the cloud, processed, and results are sent back to the end-users. To solve this problem, CISCO has introduced fog computing, wherein fog nodes filter the data to be sent to the cloud and process the rest of the data. A faster version of fog computing is edge computing, wherein edge devices do most of the computation, making the end-user experience faster and smoother. Although fog computing and edge computing are helpful in reducing the load of centralized clouds and giving a faster end-user experience, these systems involve many challenges and security concerns, which are discussed in this chapter.

Keywords

Cloud computing; fog computing; edge computing; security attacks; privacy preservation

1.1 Introduction

Data is the lifeblood of the modern age, providing significant business insight as well as real-time control over crucial business processes and activities. Businesses nowadays are flooded with data, and massive amounts of data may be routinely acquired from sensors and the Internet of Things (IoT) devices working in real time from remote locations and hostile operating environments practically from anywhere in the world. The IoT lets a device connect to the internet – a concept that has the potential to drastically alter our lives and workplaces. The IoT is predicted to grow faster than any other category of connected systems. The virtual data stream is also responsible for radically altering the way in which firms handle computing [1].

Fig. 1.1 shows the statistics about the number of IoT-connected devices worldwide in 2018, 2025, and 2030 according to the reports of Statista [2]. As per these statistics, at the end of the year 2018, there were 22 billion IoT-connected devices around the world, and by the year 2030, around 50 billion IoT devices are forecasted to be used, resulting in a massive web of interconnected devices spanning everything from smartphones to kitchen appliances.

Figure 1.1 Rising trend in IOT devices. (Source: https://www.mdpi.com/1424-8220/22/3/995)

Fully automated systems are being increasingly deployed in almost all locations to monitor and manage the associated infrastructural components, which has been possible with key technologies like the IoT and cloud computing. The cloud computing network is nothing but a set of high computing servers situated at a remote location made available for its users. IoT sensors record the different physical parameters and pass these to the cloud computing network. The cloud servers apply machine learning algorithms to the data and make powerful predictions based on which actions are taken. However, these IoT devices are constrained by limited energy as the cloud computing nodes are located far from the data source, for which delays and latencies are introduced in reporting to user queries. Much research has been carried out to overcome this limitation, with researchers proposing solutions such as edge computing and fog computing technologies. Both edge computing and fog computing make the lighter versions of cloud computing, which is easily accessible to the users with improved latencies and other features.

This chapter summarizes the computing paradigms with their architecture, advantages, drawbacks, security issues, etc. The main objective of this chapter is to set the context and provide background details that are necessary to understand the further chapters in the book.

The organization of the chapter is as follows: Section 1.2 summarizes the computing paradigms in brief and their comparison; Section 1.3 describes cloud computing in detail with the architecture, advantages, and drawbacks; Section 1.4 summarizes the fog computing paradigm with its pros and cons and security issues; Section 1.5 explains edge computing in detail with its background, necessity, architectures, advantages, drawbacks, and use cases; Section 1.6 highlights the challenges of edge computing implementation; and finally Section 1.7 concludes the chapter.

1.2 Computing paradigms: cloud, fog, and edge computing

Cloud computing provides the possibility to outsource the data processing and storage to a bigger and more powerful pool of servers with ease and per the requirement. This allows the users to have low resources on their devices. Cloud computing has been widely accepted in the industry for a variety of application needs as it facilitates the computing, storage, and network infrastructure capabilities getting extended with optimal cost savings.

Cloud computing enables businesses with a ubiquitous, convenient, and on-demand services platform. This has led to lesser investments by individuals and businesses, allowing them to focus on their goals rather than computing infrastructures.

Data is generated by IoT devices and saved in the cloud, which necessitates faster processing and responses. As a result, the closer the user is to the cloud, the faster the performance speed of the device [3]. Transmission delay increases as the distance between the cloud and the IoT device user grows. Because of the increased transmission delay, IoT device users may experience certain performance issues. Thus the traditional cloud computing architecture based on a centralized data center and the internet is not well suited for applications that require continuous transferring of the expanding stream of real-world data. Bandwidth constraints, latency concerns, and unpredictability in network interruptions can all work together to standoff such initiatives.

Fog computing, which refers to extending cloud computing to the network edge [4], was introduced by Cisco to address the performance issues in cloud computing. The architecture pushes intelligence down to the LAN level of network design, where it processes data in the IoT gateway or fog node. Simply put, it entails bringing the computers closer to the sensors they are communicating with [5]. However, the design of fog computing depends on multiple links in a communication chain in transporting data from our physical world assets into the digital world of information technology. Each of these links may be a potential point of failure [6,7].

The issues in fog computing have expanded the scope of edge computing that demands data processing at the network edge. Edge computing has been growing fast since 2014 with many applications and the potential to minimize latency and bandwidth charges, addressing the limitation of the computing capability of the cloud data center, boosting availability, and safeguarding data privacy and security [8].

Fig. 1.1 [9] depicts the pyramid structure of these paradigms in terms of the number of devices engaged at each level. It is clearly seen from this figure that billions of edge devices are interacting with millions of fog devices, which in turn are communicating with thousands of cloud data centers. The computational capacity of cloud computing is the highest, and that of edge computing is the lowest (Fig. 1.2).

Figure 1.2 Recent computing paradigms. (Source: https://www.mdpi.com/2673-4001/2/4/28)

Table 1.1 provides a comparison among the three computing paradigms.

Table 1.1

1.3 Cloud computing

According to the definition provided by the National Institute of Standards and Technology (NIST), cloud computing is a model that supports on-demand access for sharing the storage and computing infrastructure [10]. Cloud data centers are constructed with a larger collection of virtualized resources that are highly accessible and reconfigurable to the increasing workloads and support the pay-as-per-use cost model [11], making it a more convenient and preferable choice among the users. Google, Microsoft, and Amazon cloud service providers are the major giants in this domain and provide powerful cloud infrastructure for user services.

1.3.1 Architecture

The cloud architecture is as shown in Fig. 1.3, with the front end referring to client management, which acts as the graphical user interface (GUI), and the back end referring to application, service, management, runtime cloud, storage, infrastructure, and security. Each layer in the backend architecture is essential for completing the action over the cloud. The internet is the bridge to establish the connection between the front end and the back end.

• Application: It acts as the platform for the user to get services that are provided by the cloud

• Services: The cloud services are managed according to the client’s requirements, which are-

• Software as a Service (SaaS): It is a platform-independent service in which the user can go to the desired application per the requirement, for example, the Google storage system.

• Platform as a Service (PaaS): It is a platform-dependent system that provides the platform for making an application for the client, for example, OpenShift

• Infrastructure as a Service (IaaS): This is responsible for managing application data, middleware, runtime, for example, AWS

• Management: It coordinates among all the layers in the backend for communication.

• Run-time cloud: It provides the virtual environment for execution and computing.

• Storage: It provides storage infrastructure for customers.

• Infrastructure: It refers to both the software and hardware components of the cloud, including the devices for network connection and software for virtualization.

• Security: It provides confidentiality to clients’ data to prevent attacks.

Figure 1.3 The architecture of cloud computing. (Source: https://www.mdpi.com/2079-9292/10/15/1811)

1.3.2 Advantages of cloud computing

• Storage cost: Cloud computing allows clients to spend less on storage as they no longer need storage disks as the cloud provides massive storage space.

• Increased computing power: Access to cloud computing offers access to the enormous computing power of the data center since the clients are no longer limited by the capacity of the desktop computer.

• Collaboration: Cloud computing allows clients located in different locations to conveniently and securely collaborate with each other. These collaborations can be internal across departments and also externally with clients.

• Reduced software costs: Clients do not need to purchase the software as cloud computing provides access to software per the requirements. Also, cloud computing updates the software automatically, and users do not have to worry about the software.

• Scalability: Clouds are easily scalable, so clients can add and remove resources per their needs.

• Reliability and recovery: As the cloud maintains a lot of redundant resources, failures can be effectively handled. Also, it provides the most efficient recovery plan after failures.

1.3.3 Drawbacks of cloud computing

In spite of the advantages of cloud computing, it has the following drawbacks:

• Increased delays in data uploading: Cloud computing requires the data to be fetched before any data processing has started, causing delays, especially in real-time applications where data uploading takes time, due to which responses are delayed.

• Latency in user network access: If the interfaces between the user and the IoT network are hosted on the cloud, then some additional time is needed to direct the user data to the IoT network causing latencies.

• Limited customizations: As the applications and services hosted on the cloud are defined by the service level agreements (SLAs) between the service providers and customers, limited customizations are possible.

• Dependency on internet: Cloud computing requires a good internet connection, without which it is not possible for the cloud to operate and communicate with the end-user.

• Security concerns: In cloud computing, the confidential data of the customer is hosted on the remote cloud provided by the service providers. As such, there are chances that malicious cloud providers misuse the data. Also, the security of the data can be compromised by the attacker during the communication between the user and the cloud network.

• Technical support and issues: Cloud service providers are required to give 24×7 technical support for customer queries. Many service providers are making customers rely on FAQs and online help. Also, the cloud network may experience outages or other technical issues, which need to be monitored continuously and solved immediately.

1.4 Fog computing

Cloud computing has facilitated the on-demand delivery of IT resources like storage, computing power, hardware, etc. In spite of these advantages, latency is a major problem with cloud computing. All IoT devices send data to the cloud for further processing and storage [12]. The IoT environments are constrained with respect to bandwidth, processing, memory, energy, etc., as a lot of data is sent to the cloud, and a lot of energy is invested into it [13]. So, the idea of fog computing comes into the picture, which extends the cloud nearer to the IoT devices [14].

As per the statistics, 40% of the world’s data comes from sensors alone, and 90% of the world’s data are generated only during the period of the last 2 years. There are almost 250 million connected vehicles worldwide and 30 billion IoT devices [15].

The ability of the current cloud model is insufficient to handle the requirement of the IoT because of the volume of data, latency, and bandwidth. Volume refers to the huge amount of data produced by different end-user applications; latency refers to the time taken by a packet for a round trip that causes the delay, which is not acceptable for time-sensitive applications; and bandwidth refers to the bit rate during transmission wherein if all the data generated by IoT devices are sent to the cloud for storage and processing then traffic will be so heavy that it will consume almost all bandwidth