As the control center of telecom network, core network is the brain of telecom network. In the 5G era, 5G core network is the key to enable a wide range of businesses. The on-demand definition and dynamic adjustment of network slicing can meet differentiated requirements of a wide range of businesses. MEC provides advantages in high bandwidth, low latency, and near-end deployment, thus creating new service and revenue opportunities and new business models. However, this poses unprecedented challenges to the traditional manual decision-making and O&M:
- Requirements for new technologies: With the introduction of such technologies as NFV, Microservice and slicing, the NEs are decoupled vertically in layer-3 mode, the number of NFs increases, so O&M management needs to introduce the AI capability to improve the troubleshooting efficiency and implement real-time network optimization and adjustment.
- Requirements for new networks: There are multiple MEC devices at the edge, and the sites are far away, which increases the O&M manpower cost and time cost dramatically. The AI capability should be introduced to improve the deployment and O&M efficiency of the edge devices and reduce the manpower cost.
- Requirements for new services: The complex 2B service scenarios of the network bring about differentiated requirements for SLA, and demand higher network quality assurance and customer experience assurance. O&M management needs to introduce AI capability to implement active network O&M, so as to discover faults and index degradation in advance and ensure reliable network and service operation.
ZTE uSmartNet sub-solution -- uSmartNet-CN Core Network Intelligent solution, which is based on the ZTE uSmartInsight AI platform, flexibly introduces AI engine at the infrastructure layer, network layer and management control layer. It constructs a hierarchical and closed-loop intelligent O&M system, and supports the intelligentization and automation of the entire O&M process of network planning, deployment, maintenance and optimization.
Figure 2.1 ZTE uSmartNet-CN Core Network Intelligent Solution Architecture
ZTE uSmartInsight AI platform is AI application programming and training platform, which provides data processing capability, AI training capability and application reasoning capability. It supports both traditional machine learning and in-depth and federal learning. It integrates 300+ machine learning operators and 100+ application components oriented to telecom network field.
Figure 2.2 uSmartInsight AI Platform
- Hardware: Based on GPU, FPGA, CPU or HPC cluster, the platform supports general servers and blade servers.AI framework: The platform supports mainstream in-depth learning frameworks such as Tensorflow and Caffe, and machine learning frameworks such as Sklearn and Spark MLlib. It optimizes such key technologies as parallel computing acceleration and inference pruning.
- AI algorithm components: The platform covers wireless network, wired network, core network and other global algorithm models, such as alarm association model, capacity prediction model, user behavior model and traffic model.
- AI application components: The platform, which is oriented to the specific application of 5G intelligent, flexibly supports various application scenarios, such as intelligent prediction, RF fingerprint, intelligent slicing and root cause analysis.
ZTE uSmartInsight platform is based on the micro-service architecture, and its can be flexibly introduced and deployed according to the application scenarios. The product forms include the following:
- Centralized AI platform: The data, which is deployed in a centralized manner, works with the data lake or big data system to support the development and offline training of the network-wide AI model and implement cross-domain intelligence.
- Lightweight AI engine: The engine is combined rapidly with the existing O&M management and control system to support reasoning of AI model and some online training, and achieve single-domain intelligence.
- Real-time AI engine: It can embed NEs or base stations to implement the NE's high real-time strategy.
3 ZTE uSmartNet-CN Intelligent O&M Process
ZTE uSmartNet-CN introduces AI capability engine in a hierarchical mode. According to the decision-making basis output by the AI training platform, it automatically executes the management strategy, and endows the network with the capabilities of intelligent perception, modeling, commissioning, analysis and judgment, and prediction of the network. In addition, it supports the end-to-end intelligent O&M process of the core network with a series of automation tools.
Figure 3.1 ZTE uSmartNet-CN Intelligent O&M Process
Intelligent planning: Based on massive data analysis and AI prediction, uSmartNet-CN identifies valuable areas, rapidly converts service requirements into network requirements, and automatically designs network parameters and network pre-plans to improve the efficiency of network planning.
- Intelligent deployment: With the automatic deployment tool, the platform automatically completes the installation and deployment, data configuration and check at each layer of the network, intelligently matches the test scenarios and cases, and automatically tests the service functions and interfaces. The deployment period is shortened from several weeks to several days.
- Intelligent maintenance: Through 3600 panoramic detection of network health and network KPI exception perception, based on the time series algorithm, the platform can predict faults and network problems, and change passive O&M into active O&M. When a fault occurs, the fault can be rapidly delimited and located through the associated analysis of CHRs (Call History Record), alarms and logs.
- Intelligent optimization: By collecting such data as resources, load and traffic in real time, the platform can predict and evaluate user perception, service quality and network status, and dynamically adjust network resources, topology, paths and parameters to eliminate potential network problems and ensure the running of high network quality.
ZTE core network products will continue to focus on the automatic and intelligent efficiency engine based on AI and expert experience, and actively carry out joint exploration and practice with global operators to help them move towards the Zero Touch smart O&M era.