1
Introduction

The fifth-generation (5G) and the upcoming sixth-generation (6G) wireless networks are poised to be the most significant and transformative technology in the coming decade, disrupting numerous industries ranging from energy, agriculture, manufacturing to transportation, retail, healthcare, entertainment, and financial services. Surprisingly and unbeknownst to most outsiders, one of the key technology enablers is the new computer architecture and software design model, in addition to the wireless communications technology [1, 2].

The main objective of this book is to help university students and professional engineers understand the 5G/6G edge computing architecture and to describe step by step how to unleash the full 5G/6G potential using custom edge computing technologies.

This book is divided into five parts:

• Part 1 (Introduction):
  1. - Chapters 1 and 2 introduce the concept of edge and custom computing and provide an overview of 5G/6G technologies.
• Part 2 (Theory):
  1. - Chapters 3 and 4 discuss how to use high-level synthesis (HLS) and coding theory to realize secure edge acceleration with high performance.
• Part 3 (Architecture):
  1. - Chapters 5 and 6 elaborate the 5G/6G hardware and software acceleration architecture.
• Part 4 (Applications):
  1. - Chapters 7 and 8 describe a few 5G/6G killer applications with acceleration technology including some practical development strategies.
• Part 5 (Future Roadmap):
  1. - Chapter 9 discusses the road ahead in the coming decade.

1.1 Introducing 5G and Internet of Everything

With 20 times higher bandwidth and 10 times reduction in network latency, fifth-generation (5G) technologies enable many new applications, such as remote surgery and autonomous driving. These incredible applications require both high bandwidth and extremely low latency. To realize these useful applications, we need high-performance communication and computing platforms demanding different kinds of acceleration technologies, known as heterogeneous acceleration architecture (Figure 1.1).

Although the fundamental wireless technology - wireless spectral efficiency - has only increased three times from 4G/LTE to 5G, the area traffic capacity is 100 times larger. This is made possible mainly by building many more 5G cell towers, each covering a smaller area and supporting a wider spectrum. A typical 4G/LTE cell tower may have a range of 10 km, but a typical 5G tower may have a range of 500 m or less.

Figure 1.1 Comparison between 4G and 5G features and performance metrics.

The 3rd-Generation Partnership Project (3GPP) is responsible for driving the 5G telecommunications standards, which are detailed in the 3GPP Specifications Release 15 (NR Phase 1) and Release 16 (NR Phase 2) documents. The 3GPP 5G network architecture defines the network entities based on their functions and nature (control and data planes) [3-6].

There are three technical specification groups (TSGs) within the 3GPP:

• TSG SA (services and systems aspects) focuses on the overall architectures and services.
• TSG CT (core network and terminals) focuses on the core network (CN) architecture and terminal interfaces.
• TSG RAN (radio access networks) focuses on the radio transmission and its technical requirements.

The TSG SA defines three different sets of requirements for new 5G usages:

• Enhanced Mobile Broadband (eMBB): a new requirement that defines higher data rates, traffic and connection densities, and user mobility.
• Massive Machine-Type Communications (mMTC): a new requirement that supports very high traffic densities of devices.
• Ultra-Reliable Low Latency Communications (URLLC): a new requirement that provides very low latency and very high communications service availability (Figure 1.2).

Figure 1.2 compares the bandwidth and latency requirements of some 4G and 5G applications. The typical 4G network latency is 10-20 ms, while the minimum 5G network latency can go as low as 1 ms. The maximum 4G network bandwidth is around 100 Mbps, but the 5G network bandwidth can go up to 1 Gbps. Some new mobile applications, such as mobile telepresence, may only require higher network bandwidth. On the other hand, most disruptive applications, such as autonomous driving and smart energy grid, may demand very short network latencies.

Although human beings may not perceive any difference between 5 and 50 ms, this difference is so important in many industrial robotic systems. Indeed, an interconnection of machine systems - called the Internet of Everything (IoE) - can take full advantage of the three new 5G features (eMBB, mMTC, and URLLC) [7, 8] (Figure 1.3).

In 2012, Cisco Systems extended the concept of the Internet of Things (IoT) and coined the term "the Internet of Everything (IoE)," representing a networked connection of people, processes, data, and things. It goes beyond simple machine-to-machine (M2M) communications, forming a network of networks connecting all data, technologies, processes, and people [9-12].

Figure 1.2 Latency and bandwidth requirements of some typical 4G and 5G applications.

Figure 1.3 3GPP 5G network architecture.

Figure 1.4 Ubiquitous Internet of Everything (IoE) devices.

Today many IoE devices are adopting a tethered communication scheme - using Wi-Fi or Bluetooth to first connect to a smartphone or an access point - to communicate with other IoE devices. Unfortunately, tethered communication is slow, unreliable, and insecure, limiting the performance of these applications [13, 14].

The 5G/6G wireless network enables IoE devices to adopt an untethered communication scheme, thus allowing each device to have a secure channel connecting to a radio access network (RAN) with guaranteed network latency, bandwidth, and security (Figure 1.4).

According to Juniper Research, IoE devices reached 46 billion units by 2021. Figure 1.4 depicts a framework of different connected devices and things. The top IoE applications include smart factories, smart buildings, smart grids, connected gaming and entertainment, smart vehicles, remote healthcare, remote education, and connected marketing and advertisement.

These ubiquitous IoE devices have some essential characteristics:

• Resource Limitations: Many IoE devices are limited by size, weight, and power (SWaP) and do not have sufficient resources to process or store information locally. Accordingly, the information is offloaded to a remote server for storage and processing.
• Network Latency: Extremely low network latency is needed for time-critical applications. Industrial control systems can only tolerate delays of the order of milliseconds. Autonomous vehicles and virtual reality applications also have similar latency requirements. The network latency can be excessive if the servers are in a remote cloud.
• Network Bandwidth: A large IoE network can generate a huge amount of real-time data. For example, 120,000 CCTV devices transmitting 1080p video to a remote cloud may generate 1 Tbps traffic. If the servers are in a remote cloud, the 1 Tbps network data can cause serious traffic congestion in the Internet backbone.
• Security and Reliability: Many critical applications, such as energy grids and smart factories, must be ultra-secure and reliable. Due to the multi-hop nature of the Internet backbone, it is probably impossible to guarantee ultra-security and reliability with a remote cloud server.
• Reconfigurability: After deployment, an IoE system oftentimes requires future security patches and performance updates. While it is easier to update a generic software architecture, it is difficult to update a high-performance hardware architecture. This is an important reason why custom computing is desirable for 5G computing architecture.

1.2 Edge Computing Architecture

Although 5G can provide large bandwidths and low latencies, if the other end of the communication is far away, these parameters cannot be guaranteed. In particular, the communication latency is limited by the speed of light. For example, a distance of 4000 km - between Los Angeles and New York - has at least 20 ms delay on fiber optics [15-17].

In reality, the actual latencies, including all electronic and electro-optic delays, are much longer. Table 1.1 shows the typical round-trip latencies from Tokyo to some target cities.

Table 1.1 Round-trip time (RTT) from Tokyo to overseas cities.

Source: Adapted from http://Wondernetwork.com.