[Technical details] Alibaba Cloud AIoT object model support equipment scale has exceeded 100 million

Introduction to The object model technology introduced in this article, for Alibaba Cloud AIoT, the object model technology has been accumulated for many years, so it can enable various hardware products to achieve truly intelligent connections.

Object model technology is a very important technology for Internet of Things companies, because in order to realize the interconnection of all things, the precipitation of the object model system is necessary to enable various hardware products to achieve truly intelligent connections. For Alibaba Cloud AIoT, the object model technology has been accumulated for many years. At present, Alibaba Cloud AIoT object model technology has become the industry leader.

For the industry, Aliyun AIoT’s object model technology mainly has two core advantages—ecological advantages and platform advantages. Relying on Alibaba’s strong ecosystem and tens of millions of partners and products, it will bring Alibaba Here comes a unique model precipitation. As a platform-based company, Alibaba is different from vertical companies in that it can openly support customers in all walks of life. These two advantages can be called the killer features of Alibaba Cloud's AIoT object model technology. At present, the scale of supporting equipment for Aliyun's AIoT model has exceeded 100 million.

Now, Alibaba Cloud AIoT material model technology expert Xiong Yiqun has brought you a comprehensive guide to material model technology. Why is the analysis of material model technology so important?

Three issues that the object model technology needs to understand:

1. Why do we need a physical model?

Massive IoT data, equipment, services, heterogeneous equipment and data description methods are difficult to understand and interoperable. First of all, the industry chain has its own internal system, with diverse roles of modules, chips, platforms, and solution providers, and cross-role collaboration. , Data standards are different, collaboration is difficult; secondly, collected data is difficult to analyze, difficult to structure, data utilization efficiency is low, and data value is difficult to mine; finally, as industry applications and equipment volume increase, new applications need to target different equipment protocols Repeated development is difficult to scale.

2. What problems can the object model technology solve in the industry? At present, the Internet of Things industry generally has the problems of equipment islands and strong coupling between software and hardware development. It is necessary to build a unified model description language and a unified modeling for physical entities. The object model as an abstract layer of objects shields the differences in the underlying terminal and standardizes the equipment. The ability to express and interact with each other greatly reduces the cost of IoT application development and rapid replication.

3. What value can the object model bring?

• Low threshold access: Provide basic equipment modeling and interactive protocol capabilities. This is the most basic value. All devices need to be modeled and interacted in the cloud. Whether the physical model and protocol design is professional enough, this is actually the threshold for most small and medium-sized enterprises. They don't realize at the beginning, and design at will, and the disadvantages will be reflected as the scale and business changes.

• Standardization: As the abstraction layer of the Internet of Things, the object model is similar to the difference between operating system shielding hardware and JVM shielding OS. Through the ability expression and interaction methods of standardized equipment, it solves the protocol differences and software and hardware development under the severe fragmentation of the Internet of Things. Coupling, long full-link verification process, equipment islands, data islands and other issues.

• Ecologicalization: Once the software and hardware are developed and interacted based on the standardization of the object model, the multiple roles surrounding the Internet of Things, including ISV, SI, IHV, etc., can be mutually resolved in the links of equipment development, production, operation and maintenance, sales, integration, and operation. Coupling improves the circulation of equipment and promotes ecologicalization.

depth analysis

What kind of technical challenges does the physical model face now?

Take a light bulb as an example

Let's first take a look at what capabilities or features an ordinary smart light has, such as switching, color tone, brightness, over-temperature alarm, and factory resetting capabilities, including the status of sensor collection, danger alarm, and controller Executable instructions. Then the complexity and difference of equipment in different industry scenarios are different. From simple consumer equipment "lights" to complex industrial equipment "boilers" all need to be expressible. It is still very useful to define a set of sufficiently abstract and universal object models. It is challenging, so it needs to follow certain design principles, such as simplicity, universality, scalability, modularity, and ease of use.

expand, the technical challenges of the physical model include these

The physical model describes the complete capabilities of all heterogeneous equipment and plays a role in the entire life cycle of the equipment. Therefore, there are the following problems that need to be resolved during the design of the physical model:

• Universality: The definition and design of the object model can be adapted to all equipment, and it needs to cover many different industries such as industry, life, agriculture, and transportation. Therefore, in the design, it is necessary to find the most essential commonalities of the equipment and abstract a set of models.

• Super-large points and super-complex structures: Especially in industrial scenarios, it is usually necessary to digitize traditional automation systems containing a large number of sensors (ten-thousand-thousand level), which poses a very big challenge to the object model, and the complexity of the object model becomes the same as the physical entity. The complexity of the environment is positively correlated, and we need to find the most essential cracking method to prevent the complexity of the object model from becoming uncontrollable.
• Internationalization: The equipment can be sold to all parts of the world at any stage, and the physical model enables the equipment to have multi-language capabilities everywhere.
• Pluggable: The important thing about the rapid development of industrial civilization is standardization. A large number of complex devices can be standardized and assembled, such as cars, ships, homes, etc. Modules can be dynamically plugged and unplugged according to product characteristics, so the object model also needs to be able to adapt The physical device module is pluggable.
• Safety development: The object model is defined in the equipment development stage and is referenced in the equipment operation stage. It is necessary to ensure that the object model defined or debugged in the development stage does not affect the equipment running in the production stage.
• Fast debugging: Traditional hardware development and software development need to be coordinated with the entire link for debugging, which has a long cycle and high costs. With the object model, the debugging stage needs to ensure that the hardware and software are decoupled and not dependent on each other.
• High reliability: Offline is the big difference between the Internet of Things and the Internet. There are a large number of offline physical devices, geographical locations and application scenarios and their wide range. The cost of on-site operation and maintenance of equipment problems is very high, and it has a great impact on society. The reliability requirements of the physical model in the equipment operation stage are very high.
• Rollback: In order to ensure high reliability, the material model needs to be rolled back quickly in the event of an abnormality from development to operation.
• Adaptable: Because there are already many device models and interactive protocols in the industry, such as Modbus, opc, etc. in the industrial scene, ble, zigbee, etc. in the life scene, and of course there are a large number of third-party platform proprietary protocols, in order to help these stock devices to be able to use Object model, the object model needs to have model and protocol adaptation capabilities.
• Unified interaction protocol: In addition to being expressible, the device also needs to be accessible. The object model not only needs to define the device capability description specification, but also needs to define the way the device is accessed. All devices can use the same set of interactive protocols for access. The design is also There are many challenges. For example, resource-constrained equipment, weak network environment equipment, and industrial edge equipment will have different protocol requirements. Some pursue low power consumption, some pursue less bandwidth, some pursue large points and high frequencies, and some pursue more networks. Cascade and so on.
• Twin agent: The core value of the physical model lies in the digitization of physical entities. After the physical entities are digitized on the cloud, a digital twin will be built. The data model and access methods of the digital twin are based on the physical model. The digital twin acts as an agent for the interaction of physical devices with industry applications. , So as to achieve the decoupling of software and hardware. However, what features the twin agent should have to ensure that the hardware capabilities can be accessed efficiently and reliably is very challenging, such as device disconnection or abnormal conditions. Different. Of course, it is a bigger challenge for twins with massive data to guide the production, operation and maintenance of physical equipment in the opposite direction based on data intelligence, and to achieve the goal of sharing intelligence with physical equipment.

Then how should we design the object model?

In the early days, most IoT platforms such as Azure and AWS only had connectivity and basic management capabilities, and did not have digital device modeling and digital twin capabilities. However, in the past two years, almost all IoT platforms have begun to pay attention to IoT models and digital twins. Construction. Most of the device modeling uses the idea of object-oriented languages, such as WoT, OPC, OMA, OCF, CWMP, AllJoin, etc. The abstract ability of object-oriented languages has been proven in the decades of computer programming development. The model definition is also fully used for reference, but it is different due to the Internet of Things.

We use class in the object-oriented language java as an analogy. Class uses attributes and methods to describe the state and behavior of objects; the object model can also use attributes and methods to describe the state and behavior of objects. At the same time, in combination with the characteristics of the equipment, we have extended the schema of the object model to a certain extent, which is defined as the three elements of attributes, services (methods) and events. Events are a special type of attribute, such as air-conditioning failure alarms, which are of high severity. Real-time is strong, generally need to monitor and respond in time. In order to describe the equipment more accurately, the object model also defines very rigorous data specifications for each data type. For example, in addition to the data type, it is also necessary to define specifications such as data range, accuracy, and step length.

[The picture shows the basic schema of the object model (does not include a series of high-level features such as modularity, multi-language, and multi-version)]

solving these challenges, the technical architecture of the object model is presented

In addition to describing the capabilities of physical entities through the three elements of attributes, events, and services, the Aliyun AIoT object model also supports thousands of large points, multi-language, multi-version, multi-module, multi-level cascade, protocol adaptation, Cloud-side-end integration and other capabilities can meet the definition demands of different scenarios such as life, city, and industry. Of course, in order to cope with the series of technical challenges mentioned above, we have also built a set of digital capabilities for physical entities by building the Alink protocol and digital twins.

more thing to note, the object model and data standards are different.

The object model can describe the capabilities of a device with the same set of schemas, but due to the fragmentation of the Internet of Things, everyone's definition of device capabilities is very different. For the same air conditioner, the capabilities defined by different manufacturers will be different. It is equivalent to the standardization of the interface in the object-oriented language, but the implementation is not standardized. The core of the data standard is to reduce differentiation.

The data standard is a batch of standardized materials that can be used for the assembly model. The material model construction process can easily select materials from the data standard library for building block construction.

In the case of severe fragmentation in the traditional field, it is very challenging to define data standards, which can usually only be defined by deep ploughing in traditional industries. Therefore, we are more to introduce these industry leaders to contribute data standards than to develop ourselves. The precipitation of the Alibaba Cloud IoT data standard mainly comes from the ICA Standards Alliance. The ICA standard library includes three types of materials: basic resources, functional modules, and object templates:

• Resource: The most atomic capability in the standard library, with three types (three elements): attributes, events, and services;

• Function module: a collection of a set of resources. The resources in the collection can be an assembly of existing resources in the standard library, or a new resource added to the current function module;
• Object template: a collection of a set of functional modules and a set of resources. The modules and resources in the collection can be an assembly of existing modules and resources in the standard library, or they can be newly added resources in the current object template;

following figure describes the relationship between the object model and the data standard

Finally, we look at the schematic diagram of the bulb object model

What is the core value of the data standard?

The ICA standards alliance led by Ali has deposited a batch of standardized data models. The core value is that the modeling process can be quickly assembled, built in building blocks, and improved modeling efficiency; on the other hand, standard object templates can promote the standardization of software and hardware. In this way, software vendors and integrators can plug and play the purchased hardware.

Copyright Statement: content of this article is contributed spontaneously by Alibaba Cloud real-name registered users, and the copyright belongs to the original author. The Alibaba Cloud Developer Community does not own its copyright and does not assume corresponding legal responsibilities. For specific rules, please refer to the "Alibaba Cloud Developer Community User Service Agreement" and the "Alibaba Cloud Developer Community Intellectual Property Protection Guidelines". If you find suspected plagiarism in this community, fill in the infringement complaint form to report it. Once verified, the community will immediately delete the suspected infringing content.