— Xinghe Network Large Model, Powering the iMaster NCE Network Digital Map

[Shanghai, China, September 21, 2023] At HUAWEI CONNECT 2023, Huawei released the industry’s first large network model — Xinghe Network large model. This future-oriented model empowers the iMaster NCE network digital map, and accelerates network digitalization and industry intelligence.

Leon Wang, President of Huawei Data Communication Product Line, introducing Huawei Xinghe Network large model

Huawei Xinghe Network large model is a communication network model trained based on Huawei PanGu models. The large model aggregates more than 20 billion corpuses (such as configuration items and command lines) in the Huawei data communication domain and the experience of more than 30,000 network experts. It has powerful semantic understanding capabilities, and supports knowledge Q&As, interactive service analysis, and decision-making assistance.

Huawei Xinghe Network large model stands out with three highlights:

  • Multi-dimensional awareness: Detects multi-dimensional network data and application data in real time and completes hierarchical modeling.
  • Proactive prediction: Establishes a prediction model to accurately predict the network traffic trends, network and security risks, and impact scope.
  • Interactive dialog: Delivers unparalleled human-machine interaction experience. Leveraging the network digital map and large model dialogs, Xinghe Network allows users to get access to a wider range of service scenarios.
  • Decision-making assistance: Sets up a hierarchical multi-objective decision-making model to intelligently recommend the optimal handling policy and suggestions.

Huawei Xinghe Network large model has three key benefits.

  • Multi-Dimensional Scanning, Identifying Network Risks in Advance

The model performs hierarchical modeling on parameters in approximately 10,000 dimensions (such as networks, applications, and devices) and implements in-depth correlation inference of abnormal data to predict network faults and risks in real time and automatically recommend optimization policies.

  • Dual-Model-Driven, Creating Ultimate Network Experience

The application experience prediction model is used to evaluate the application health from six dimensions and predict application experience deterioration. The application-based traffic steering decision-making model automatically learns service level agreement (SLA) requirements of applications and intelligently recommends traffic steering policies based on link loads and tariff packages, thereby achieving optimal application experience on the entire network and minimizing private line costs.

  • Multi-Dimensional Modeling and Analysis, Predicting Unknown Threats in Advance

Huawei Xinghe Network large model performs modeling and analysis from multiple dimensions (such as identify, terminal, and session), identifies abnormal traffic with higher accuracy, analyzes the impact scope of network faults, and traces the attack path to prevent network security risks in advance. Beyond that, the model can automatically generate interpretable handling solutions or virtual patches for any network attack, enabling dialog-like security operations. All of this makes network security threats manageable and controllable.

Huawei iMaster NCE network digital map is the industry’s first product and solution to integrate Huawei Xinghe Network large model. Oriented to the campus network, wide area network, and data center network scenarios, the digital map enables network-wide unified management and functions as a digital brain for enterprise networks, accelerating the intelligence of a wide range of industries.