The use of edge computing in factory environments is not new. Programmable logic controllers (PLCs), microcontrollers, servers and PCs for local data processing, and even micro data centers are edge technologies that have been present in factory systems for decades. The kanban system, punch card system, historical database, and workshop management software seen in the workshop can all be considered as part of edge computing. For many years, we also didn't feel like they were edge devices.
Origin
So, why do people come up with the concept of edge computing? In fact, edge computing was proposed after the emergence of cloud computing. In Internet applications, various application software are hosted to remote data processing centers, forming a cloud computing service model. SaaS, PaaS services are proposed. Small and medium-sized enterprises no longer set up an internal computing center as in the past. In order to realize cloud services, they have developed a series of platform architectures, including new technologies such as virtual machine clusters, containers, and microservices. Among them, containers and microservices are the most important cloud computing technologies. They establish a relatively independent operating space for applications, and can use independent operating systems and operating environments (such as various libraries) inside. As long as the compiled application image (images) is deployed to a container in the cloud, it can run independently. The container management system enables rapid deployment, iteration and maintenance. In Internet applications, container technology is quite mature. When cloud services are extended to "things", the "Internet of Things" appears. As a result, protocols such as MQTT have become popular. Large cloud service providers have built IoT Hub access services in cloud services. Remote devices can access the cloud IoT Hub through the gateway, and realize data storage, dashboard and control in the cloud. This is a typical IoT application system.
When the Internet of Things technology enters the manufacturing industry, the Internet of Things based on cloud service technology encounters challenges. Thousands of sensors and actuators may be installed in a workshop, and their real-time, certainty and reliability of data transmission are crucial. The cloud platform based on the Internet cannot be guaranteed, and people's natural idea is to move the technology of cloud services to the vicinity of the data source. So the edge service appeared. To be precise, the edge service is the localized deployment of cloud services, and the edge is the edge of the "cloud". Imaginative IT talents even proposed "fog computing". All of these are just concepts derived from cloud services. The core is the container, microservice and message system in the cloud service.
Advantages of edge computing
Deploying edge servers near data sources makes sense:
Meet the network bandwidth and ensure the real-time, deterministic and reliability of the industrial network
A high-performance, high-reliability network architecture can be built inside the factory to ensure real-time and reliability of data transmission. For example, time-sensitive network technology can be used.
Make the factory software containerized and micro-serviced
The micro-service of software is conducive to software modularization and app-based, and can realize agile development, rapid deployment and iteration of software. At present, MES, ERP and other software in factory management are developed in the form of microservices.
software redundancy
Another advantage of running application software in the container of the edge server is software redundancy. Software can be deployed in multiple edge servers. When a certain software fails, it can be easily switched to run in the container of another edge server. .
hardware upgrade
When the hardware performance of the edge server cannot meet the new requirements, the edge server can be easily upgraded and the original microservices can be migrated.
Industry edge computing
Edge computing is preceded by "industry", which means the following two things:
- Microservices of edge computing exchange data with industrial field devices
The industrial field continues to introduce big data and artificial intelligence applications, such as equipment health detection, image recognition, process optimization, etc. These applications are more implemented using IT technology and require a certain amount of computing power. In most cases, these new applications will not be implemented on traditional PLCs, but deployed on edge devices.
- Microservices of edge computing replace some control functions of industrial field devices
Concepts such as cloud PLC and software PLC (SoftPLC) proposed by radical IT professionals. At the same time, it is predicted that industrial control will shift from hardware-centric to software-centric. Fully software-based, standardized equipment that compresses the field equipment to the minimum function. Experts in the OT field worry about whether the reliability, real-time performance, and certainty of microservices in edge devices can be guaranteed. In the author's opinion, these problems can be solved. To give an example of the telecommunications network, the previous telephone program-controlled switches were mainly based on circuit switching, but now they are completely based on softswitching. The reliability and real-time requirements of the telecommunication network are also very strict. It takes a long time to change concepts and usage habits.
Embodied Architecture for Industrial Edge Computing
The industrial cloud edge control architecture is shown in the following figure:
The Key to Realizing Industrial Edge Computing
As mentioned earlier, edge computing is derived from cloud services. In fact, organizations that provide cloud services are trying their best to promote edge computing. By sinking into the manufacturing workshop, edge computing services are provided, accompanied by private 5G Network promotion. But they encountered the same problems as promoting cloud services. Difficult to push the boundaries of OT.
In the author's opinion, the biggest feature of industrial edge computing should be its industrial attributes, which meet OT's requirements for reliability, security, and certainty.
Industrial Attributes of Industrial Edge Services
In contrast, most edge computing architectures, technologies, and standards currently on the market are more replicas of IT cloud service architectures, which are nothing more than common technologies and routines such as containers, microservices, and message systems. Lack of support for OT. To put it bluntly, for the automation industry, "there is no service in the container!". What runs in many edge servers is still the traditional MES custom software, just with a different vest. Providing key basic services for the field of industrial automation is the key to the success of edge computing. Industrial edge computing is surnamed "industry" and should have industrial characteristics.
Perhaps people in the industry still remember GE's ambitious predix project, which also adopted container technology. But it failed due to not having enough microservices. Harting's development of the Embedded Microservices Controller didn't make much of an impact either. The author has also developed a container-based embedded controller. The profound experience in my heart is that enough microservices are the key to the success or failure of edge servers. This is the same as the function block library in the OT industry.
Microservices in industrial edge servers include three broad categories:
Basic Microservices
Data access, aggregation server microservices (DataHub, Data Aggregate Server):
Collect all on-site data and become a unified entrance for other microservices to access on-site data.
Data Storage Microservice (data Historized Database Access)
Realize time-series database storage access to field data and events
Information Model Server (Information Model Server)
Transform the unstructured data of the industrial site into the Opc UA information model, and other microservices access the field data through the Opc Ua protocol and model.
Data Dictionary Microservice
Generic Microservices
Data Visualization Microservices
Support live Kanban, web access and remote App
Data Analysis Microservices
AI Microservices
Cloud access to microservices
Knowledge base, model base microservices
Microservices related to Industry 4.0
Machine job scheduling, process parameter delivery, and industrial 4.0 machine collaboration and interoperability.
Application Microservices
MES
digital twin
digital simulation
Microservices of industrial software
As mentioned earlier, edge computing in the IT field is derived from cloud service computing in the IT industry. It is based on container technology and computer communication protocols. Compared with the information technology advocated by the OT industry, the information exchange in the Internet industry is based on communication protocols. Whether it is TCP/IP, HTTP, webSocket, or MQTT and other message systems, they lack semantic consistency and it is difficult to realize communication between machines in the OT industry. Therefore, it is difficult for microservices built based on Internet protocols to ensure the interoperability of microservices developed by different vendors. Interconnection with OT field devices is even more difficult. Some manufacturers and institutions have worked hard to develop object model libraries (mostly based on JSON) based on their own experience and technology. And these rules are written into various national standards. In the author's opinion, due to the lack of industry experience in the industrial field, it is difficult to reach a "consensus" on the developed models and standards and form an ecosystem. Standards with too many options are difficult to achieve compatibility, difficult to be adopted by most manufacturers, and have little practical and guiding significance.
In the author's opinion, microservices in industrial edge computing servers should follow the model-based engineering design ideas advocated by the OT industry, and use widely recognized technical standards in the OT industry (such as OPCUA) to construct OT information models. And the OPC UA protocol is used to realize the information exchange between microservices.
Using the model-based engineering design method advocated by the OT industry is also to realize the "low code" of microservices, such as IEC61499, IEC61131-3 and other graphic-based programming methods. And automatic code generation technology based on information model.
Based on rich experience in the industrial field, industry experts in the OT industry have formulated a large number of standards for Industry 4.0, development automation, and digital transformation. Much of the software that implements these standards is impossible to implement in traditional PLCs, and they will run as microservices in edge computing servers.
People who are committed to industrial edge computing should not stay in the IT industry comfort zone of microservices and message systems, and it may be a feasible way to develop in the direction indicated by OT industry standards.
conclusion
Using IT cloud services and edge service technologies to implement industrial software that complies with OT industry specifications is a feasible way to realize industrial edge computing. Providing a variety of microservices that support new technologies in OT is the key to success. New technologies in the industrial field will be provided to customers in the form of microservices. Ultimately, the goal of achieving the same goal and integrating IT and OT will be achieved.