A server failure prediction and processing method and system
By performing dynamic topology-aware parsing and time-series alignment on multi-source server operation logs, a time-series causal graph is constructed, and high-dimensional causal features are extracted using graph neural networks. Combined with meta-learning networks and knowledge graphs for fault prediction and processing, this solves the problems of inconsistent log parsing and reliance on human experience in existing technologies. It achieves accurate prediction and automated processing of server faults, improving operational efficiency and system reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 青岛永弘达信息产业有限公司
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for server fault prediction and handling suffer from problems such as lack of semantic consistency and causal correlation in log parsing results, limited feature representation capabilities, reliance on human experience in processing solutions, and issues with response lag and uncontrollable risks.
By collecting multi-source operation logs in real time for dynamic topology-aware analysis and time-series alignment, a time-series causal graph is constructed and high-dimensional causal features are extracted using graph neural networks. Fault prediction is performed by combining a meta-learning network with a task-adaptive strategy, program retrieval is carried out using a knowledge graph, and pre-run verification is performed using a digital twin sandbox.
It enables accurate prediction and automated intelligent handling of server failures, improves operational efficiency, reduces manual intervention costs, and ensures high availability of business systems.
Smart Images
Figure CN122412239A_ABST