Cloud data intelligent processing system oriented to multiple nodes

By constructing a directed acyclic graph and combining topology-aware scheduling and graph neural networks, data processing tasks are dynamically scheduled to nodes with sufficient link margin and idle computing resources. This solves the problems of link congestion and node idleness in multi-node cloud data processing systems, and achieves efficient data stream transmission and processing.

CN122317080APending Publication Date: 2026-06-30XIAN BANGQING NETWORK TECHNOLOGY CO LTD
6 Cites 0 Cited by

Patent Information

Application Number
CN202610756487.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-29
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

In existing technologies, cloud-based intelligent data processing systems for multiple nodes fail to effectively combine the real-time bandwidth margin of the link between the source node and the target node with the queue depth of the switch port when allocating tasks. This results in local link congestion on fixed paths for highly correlated data streams, increasing transmission latency and causing idle node computing power.

Method used

A data preprocessing device is used to extract the temporal dependency features and logical association features between data blocks, generate a directed acyclic graph, and combine it with a topology-aware scheduling engine to obtain the link bandwidth margin and congestion status. A graph neural network device is used to jointly encode the node computing load and edge link status, output the task allocation strategy, and dynamically migrate to the alternative path node during the execution of the data processing task.

Benefits of technology

Dynamic scheduling and migration eliminate the disconnect between task allocation and the underlying physical network transmission capabilities, reduce the latency of cross-switch cascading of highly correlated data streams, avoid node computing power idleness caused by local link congestion, and ensure the continuous execution of intelligent processing tasks for long-series data streams.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of electric digital data processing and digital information transmission, in particular to a cloud data intelligent processing system oriented to multiple nodes. The system comprises a data preprocessing device, a topology perception scheduling engine, a graph neural network device and a task execution device. The data preprocessing device extracts time sequence dependence and logic association of the input data stream to generate a directed acyclic graph; the topology sensing scheduling engine maps the graph to a physical network topology and calculates a link bandwidth margin and a congestion state; the graph neural network device calculates a load and side link state joint coding output task allocation strategy for the nodes; and the task execution device schedules the strong dependency task to an adjacent node of which the link margin meets a threshold value and the computing power is idle according to a strategy, and dynamically migrates the task to an alternative path node according to the link packet loss rate during execution. According to the invention, the data transmission delay of cross-switch cascade among multiple nodes is reduced, and idle node computing power caused by local link congestion is avoided.
Need to check novelty before this filing date? Find Prior Art