What Is Intent-driven Deployment?
Intent-driven deployment integrates Huawei's network planning experience in various service scenarios to build a network expert knowledge base. After a customer inputs basic network intents, a network solution that best suits the intents will be automatically generated. In addition, the system is capable of self-learning, that is, it can feed the solution modifications made by customers to the AI module for learning. This helps the system adapt to the complex and fast-changing network environment. As such, intent-driven deployment greatly improves the deployment efficiency, reduces learning costs, and minimizes manual configuration errors.
Why Do We Need Intent-driven Deployment?
- Prepare a technical proposal based on the customer's network requirements (concerned with customer flows, access quantity, and areas) and complete the network design, including selecting suitable device models, determining the networking model, and planning services.
- After the determining the solution, carry out low level design (LLD), including device name planning, topology planning, VLAN planning, device link design, IP address planning, WLAN design, and construction guidance.
- Onboard devices in plug-and-play mode and then configure VLANs, interfaces, subnets, and WLAN services for them one by one on iMaster NCE-Campus.
In this process, customers have basic network plans in the solution design stage. However, they still need to perform all the required configurations in the deployment stage, resulting in repeated workload, because the technical proposal, LLD document, and iMaster NCE-Campus are independent of each other. This is where intent-driven deployment comes in.
Intent-driven deployment provides intent conversion for customers: It generates professional network solutions based on service-related and network-irrelevant requirements entered by customers and directly guide customers to complete network configurations, slashing learning costs. In addition, the system is capable of self-learning, that is, it can feed the solution modifications made by customers to the AI module for learning. This helps the system adapt to the complex and fast-changing network environment.
ESN-free deployment is another advantage of intent-driven deployment. With ESN-free deployment, customers no longer need to accurately record equipment serial numbers (ESNs) of all devices, and only need to perform simple assembly according to the device models in the construction guide. Devices on the entire network can go online as long as the root device is onboarded. On the basis of device plug-and-play, after a device downstream from the root device is powered on, it transmits its ESN to the root device through Link Layer Discovery Protocol (LLDP). Through the root device, iMaster NCE-Campus obtains the link and ESN information of this downstream device, and fills in its ESN based on the link information available in the controller system.
How Does Intent-driven Deployment Work?
Intent-driven deployment streamlines pre-sales solution communication and order placement reference, as well as post-sales construction and network service deployment, ensuring solution consistency throughout the process. It saves repeated communication and improves service deployment efficiency. Intent-driven deployment is implemented as follows:
- Based on the customer's network requirements, the system recommends a preset or an AI-learned standard network solution, which contains networking, device selection, service network configuration, and IP address planning, and supports customization.
- The system automatically generates a technical proposal based on the network solution, which streamlines the pre-sales solution communication process.
- The system automatically generates an LLD document, which streamlines the construction process.
- The system automatically generates a solution package that can be directly imported with one click to complete all service deployment, which streamlines the service deployment process.
The following figure shows the overall solution process. First, the customer enters intents on the cloud. For example, the customer hopes to deploy a network for a supermarket with an area of about 200 square meters, and the network services include cashier and video security. Then, the cloud parses the intents into a network solution for customer confirmation. For example, one switch and five APs are required, and a cashier network and a security protection network are planned. The customer can modify the network topology and configuration. The modified result will be incorporated into data fitting in the auto-learning process, which improves the accuracy of subsequent solution recommendation. Finally, based on the solution content, the system generates three items for the supermarket: a technical proposal that explains the principles and basis of the solution, an LLD document that provides guidance on live-network deployment and device assembly, and a deployment solution package used for automatic solution execution.
Intent-driven deployment process
What Are the Key Technologies of Intent-driven Deployment?
In the process of the intent-driven deployment solution, intent parsing and recommendation stages use the intent translation technology, result modification and feedback learning stages use the AI self-learning technology, the automation stage uses automated deployment, and the device stage uses ESN-free deployment for device discovery.
Intent Translation
- Low skill requirements: Customers only need to describe their network requirements using natural language, with no need to focus on technical details of the network.
- Appropriate solution recommendation: Multiple suitable network solutions are recommended based on the built-in expert knowledge base, instead of relying on experience of planning personnel.
- Efficient network design: Network planning can be completed within seconds.
The following figure shows the principles of intent translation. Intent translation implements intent recommendation and network configuration conversion through bidirectional model binding. The system has built-in network models generated based on big data learning and expert knowledge base. The system analyzes specific industry scenarios to obtain accurate mapping between the service models and network models. Based on the mapping, the system generates network configurations by analyzing the scenario-specific service requirements entered by customers.
Principles of intent translation
AI Self-Learning
- Experience: The industry data and expert knowledge base are collected to provide experience-based preset network models.
- AI self-learning: As the network evolves, the experience-based network models may fail to keep up with customers' ever-changing service requirements. AI self-learning fully addresses this issue by continuously optimizing the network recommendation mechanism during intent-driven deployment.
AI self-learning includes the following process: When the system recommends a network solution, the customer can check the feasibility of the solution. If the solution is improper, the customer can modify it. The modified result will be recorded and be incorporated into the data fitting as new data to obtain a new conversion function. The ideal curve is continuously adjusted based on the new data points. In this way, more data points are close to the ideal curve, improving the translation accuracy.
For example, if a function relationship F exists between the area (A) of a supermarket and the number of electronic shelf labels (N) in the supermarket, the number of electronic shelf labels can be calculated using the formula F(A) = N. Then the number of APs with IoT cards can be calculated. With the increase of data samples in the supermarket (more data points are added), an ideal function F2 (ideal curve) is obtained, and the calculation result of this function is more accurate.
Automated Deployment
After intents are translated on the cloud, the system generates a solution package which contains all configurations required for network deployment. To implement automated deployment, the customer needs to import the solution package to iMaster NCE-Campus. iMaster NCE-Campus then automatically generates configuration data and delivers the configuration to target devices after the devices go online. Network deployment is then completed.
With automated deployment, the following processes can be automatically completed with one click: site creation, wireless service creation, device addition, VLAN creation, interface configuration, subnet creation, device registration, and more. The only manual operation required is to set information such as site names and passwords.
Originally, the same network services must be deployed through hundreds of operations on dozens of pages of iMaster NCE-Campus. In comparison, automated deployment in the intent-driven deployment solution requires just one click, greatly improving efficiency and eliminating manual misoperations.
ESN-free Deployment
Devices to register with iMaster NCE-Campus must have their ESNs entered on iMaster NCE-Campus, as a way to prevent unauthorized device access. If device ESNs are entered one by one, especially when there are a large number of devices, the workload is heavy and the efficiency is low.
With ESN-free deployment, iMaster NCE-Campus can automatically fill in ESNs for devices. On the basis of device plug-and-play, after a device downstream from the root device is powered on, it transmits its ESN to the root device through LLDP. Through the root device, iMaster NCE-Campus obtains the link and ESN information of this downstream device, and fills in its ESN based on the link information available in the controller system.
- Author： Zhu Yue
- Updated on： 2023-11-28
- Views： 1169
- Average rating：
