Spin-leaf network congestion control method, node, system, and storage medium

A spine node, comprising:

• at least one processor;
• and
• one or more memories coupled to the at least one processor and storing instructions for execution by the at least one processor, wherein execution of the instructions by the at least one processor cause the spine node to:
• receive network information sent by at least one leaf node, wherein the network information comprises network topology information of the at least one leaf node and a network performance indicator of the at least one leaf node;
• network the at least one leaf node and the spine node based on the network topology information of the at least one leaf node, to obtain a combined network topology, wherein
• the combined network topology is a global network topology of a spine-leaf network system;
• call a scene recognition model to process a target performance indicator, to obtain a network traffic scene corresponding to the at least one leaf node, wherein the target performance indicator is the network performance indicator of the at least one leaf node or a network performance indicator obtained by preprocessing the network performance indicator of the at least one leaf node, and the scene recognition model is obtained through training in advance based on an initial performance indicator and an initial traffic scene of each node in the spine-leaf network system;
• and perform network congestion control on the at least one leaf node based on the network traffic scene;
• obtain a scene control table, wherein the scene control table records a mapping relationship between a scene and congestion control information, and the mapping relationship is that the congestion control information corresponds to at least one scene;
• query the scene control table for target control information corresponding to the network traffic scene;
• and
• perform network congestion control on the at least one leaf node based on the target control information.

A spine-leaf network system, comprising a spine node and at least one leaf node that communicates with the spine node, wherein

• the at least one leaf node is configured to send network information of the at least one leaf node to the spine node, wherein the network information of the at least one leaf node comprises network topology information of the at least one leaf node and a network performance indicator of the at least one leaf node;
• and
• the spine node is configured to:
• receive the network information sent by the at least one leaf node, network the at least one leaf node and the spine node based on the network topology information of the at least one leaf node, to obtain a combined network topology, wherein the combined network topology is a global network topology of the spine-leaf network system;
• call a scene recognition model to process a target performance indicator, to obtain a network traffic scene corresponding to the at least one leaf node, wherein the target performance indicator is the network performance indicator of the at least one leaf node or a network performance indicator obtained by preprocessing the network performance indicator of the at least one leaf node, and the scene recognition model is obtained through training in advance based on an initial performance indicator and an initial traffic scene of each node in the spine-leaf network system;
• obtain a scene control table, wherein the scene control table records a mapping relationship between a scene and congestion control information, and the mapping relationship is that the congestion control information corresponds to at least one scene;
• query the scene control table for target control information corresponding to the network traffic scene;
• and
• perform network congestion control on the at least one leaf node based on the target control information.