Analysis of Key Technologies in Edge Computing

Edge computing integrates technologies from IT, CT, and OT fields, with MEC technology being a representative of the CT field, primarily addressing network connectivity issues. However, in addition to this, edge computing also involves technologies from the IT and OT domains, and MEC technology must work in conjunction with these technologies to enable the practical application of edge computing in operator networks.

Below, we will introduce other key technologies involved in edge computing.Figure 1 describes some of the technologies involved in edge computing through a layered approach.

Figure 1: Key Technologies in Edge Computing

Network layer technologies are designed to build a deterministic network that provides high reliability, low latency, and high security isolation. Technologies such as mobile public networks, TSN, SDN, OPC UA, and WiFi are all aimed at meeting these requirements.

Since edge computing is an extension of cloud computing, many technologies used in cloud computing can also be applied to edge computing, such as containers and microservices, although they will develop towards lightweight usage. Migrating cloud-native technologies to the edge through unified standards can promote edge-cloud collaboration, providing edge systems with functionalities and experiences consistent with those in the cloud, and enabling rapid application distribution to solve issues such as unified delivery, operation, and management of massive edge devices.

Edge computing also has unique characteristics compared to cloud computing, such as edge intelligence. By utilizing dedicated chips, hardware, operating systems, and edge intelligence technologies on the edge side, it fully leverages the heterogeneous resources and real-time responsiveness of the edge to meet the demands for agile connectivity, real-time reliability, data optimization, and application intelligence across various application scenarios.

Next, we will detail several important technologies.

1. TSN & OPC UA

In traditional Ethernet, when two workstations collide, they must wait a certain period before retransmitting the message. In the event of congestion, some messages may be delayed for a long time, leading to uncertainty in communication time. However, in industrial environments, the requirements for latency are very high, and it is essential to ensure the real-time and determinism of data transmission. TSN (Time-Sensitive Networking) is based on the IEEE802.1 standard framework and enhances the reliability and determinism of data transmission and acquisition in network communications.

TSN provides standards for the data link layer in communication protocol models. To achieve true interoperability between networks, a common data parsing mechanism is also needed, which is provided by OPC UA. OPC UA can be understood as a high-level descriptive language that uses object-oriented technology to describe physical devices, sensors, and motors as individual objects, forming a digital and structured model that allows different software to control devices as if calling objects.

TSN addresses issues at layers 1-4 of the OSI reference model, focusing on data acquisition, while OPC UA addresses layers 5-7, focusing on semantic parsing. Together, they enable interconnectivity between edge devices and networks in industrial environments, providing better network support for edge computing.

Additionally, after R16, 3GPP has increased support for TSN in IoT for industrial scenarios, enabling end-to-end TSN.

2. Container Technology

Container technology, like virtual machines, is a mode of application execution within computing platforms, meaning that if multiple applications are running on a single host, this describes how those applications operate.The execution model of applications has gradually evolved from physical servers to virtual machines and containers.

Physical servers, due to exclusive access to computing resources, can lead to significant resource waste, and the tight coupling of hardware and software can make applications difficult to migrate and scale. Compared to physical servers, both virtual machines and containers can save resources and improve operational efficiency.

Virtual machines virtualize a single computer into multiple logical computers, each of which can run different operating systems, allowing each application to run in an independent space without affecting one another, thus significantly improving efficiency, security, and portability. However, virtual machines still have high resource usage, as shown in the comparison in Figure 2. In the virtual machine architecture, each application has its own operating system in addition to the host operating system, while containers run three applications on the same host operating system. Therefore, containers consume fewer resources, making them lighter, safer, and more portable, which is very suitable for edge computing scenarios.

Figure 2: Comparison of Virtual Machine and Container Protocol Stacks

Due to the diversity of development tools and programming languages for edge computing, edge computing platforms must have secure isolation measures to ensure the normal operation of programs. Implementing resource isolation through container technology can reduce additional overhead on CPU, memory, and storage, while also enabling rapid lifecycle management of containers, allowing for container startup and shutdown at millisecond speeds.

In fact, container technology has become the standard technology for edge computing platforms, with major cloud computing vendors choosing container technology to build the underlying technology stack for edge computing platforms.

3. Microservices Architecture

Microservices architecture is a technology for the application layer in the edge computing era. During development, applications are broken down into smaller modules that are developed as independent processes. These smaller module patterns can be combined according to actual application scenarios, making them more flexible.

Microservices architecture aligns well with the needs of massive connections, complex application scenarios, and distributed deployments in edge computing, enhancing resource utilization, promoting function decoupling and reuse, and accelerating application development and innovation.

4. Edge Intelligence + Edge-Cloud Collaboration

In the early stages of AI development, due to the generally limited computing power of edge/terminal devices, model training and inference were mostly completed on cloud servers. With the development of AI chips, many chips capable of performing inference at the edge are now being applied in commercial products. For example, AI cameras and smart speakers use inference models trained in the cloud directly on edge devices.

Furthermore, edge and cloud can collaborate to continuously optimize models. Data collected at the edge, after preprocessing, is uploaded to the cloud for training. The models trained in the cloud are then fed back to the edge for further optimization of inference. For instance, Huawei’s software-defined cameras have the capability for continuous model optimization. Microsoft, Google, and Amazon also have many edge intelligence technologies suitable for industrial scenarios that can push cloud models to run on edge devices.

Currently, the most commonly applied AI capabilities include computer vision, speech technology, natural language processing, and knowledge graphs, with computer vision technology being the most utilized in edge computing applications. Applications such as smart campuses, smart factories, smart homes, AR/VR, IoT, smart security, and smart healthcare all rely on computer vision technology for environment detection, product inspection, or image processing.

As edge computing technology continues to develop and improve, new technologies will be added in the future, which is worth keeping an eye on.

Author Bio

Xu Muhong, Master of Engineering, Senior Engineer at the TIER Terminal Laboratory of the China Academy of Information and Communications Technology, mainly engaged in research on new technologies for intelligent terminals.

Contact:xumuhong@caict.ac.cn

Recommended Reading

How 5G MEC Empowers Smart Security?

Related posts

Leave a Comment Cancel reply