LLM Fault Diagnosis Using Knowledge Retrieval for Service Failures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault diagnosis methods in cloud computing and other scenarios are inefficient, costly, and lack accuracy due to reliance on manual diagnosis, data-driven or model-based approaches that require large amounts of historical data and manual labeling, and are limited to specific scenarios, failing to adapt to changes in system configurations.
Innovation Solution
Utilize a large language model (LLM) to perform fault analysis by acquiring abnormal information in natural language, retrieving similar known fault knowledge from a knowledge base, and constructing a fault analysis instruction, enabling real-time and accurate fault diagnosis without pre-training or tuning, and incorporating a reflection mechanism to mitigate hallucinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual fault diagnosis is performed, then diagnostic accuracy can be maintained through expert knowledge, but diagnosis cost and time consumption increase significantly
Solution Approach 1:
The patent introduces a language model as an intermediary between fault symptoms and diagnostic conclusions. The language model processes natural language descriptions of fault symptoms and outputs diagnostic results, acting as a mediator that automates the expert reasoning process while maintaining diagnostic accuracy. This resolves the contradiction by providing automated processing (reducing time) while preserving expert knowledge-based reasoning (maintaining accuracy).
Solution Approach 2:
The patent replaces the mechanical process of manual expert diagnosis with an automated language model system. Instead of relying on human experts to manually analyze fault symptoms and generate diagnoses, the system uses natural language processing and generative AI to automate the entire diagnostic workflow, significantly reducing time consumption while maintaining diagnostic quality through the model's training on expert knowledge.
2Productivity
If data-driven or model-based automatic fault diagnosis is used, then diagnosis efficiency improves, but the system requires large amounts of historical data and manual labeling, increasing complexity and cost
Solution Approach 1:
The patent employs a language model that can be rapidly deployed and replaced without requiring extensive historical data infrastructure. Instead of building complex, data-intensive diagnostic systems that require years of data collection and labeling, the approach uses a more flexible, easier-to-deploy language model that achieves high efficiency with minimal data preparation overhead, effectively replacing complex persistent systems with simpler, more adaptable solutions.
Solution Approach 2:
The language model serves multiple functions: it processes natural language fault descriptions, performs diagnostic reasoning, generates diagnostic conclusions, and can adapt to different fault scenarios without requiring separate models for each case. This multi-functionality reduces system complexity compared to specialized data-driven models that would be needed for different fault types, while maintaining high diagnostic efficiency.
3Speed
If traditional automatic fault diagnosis models are deployed, then real-time diagnosis is achieved, but the models cannot adapt to changes in system configurations without retraining
Solution Approach 1:
The patent implements a dynamic diagnostic system using a language model that can adapt to changing system configurations in real-time. Unlike static traditional models that require retraining when system configurations change, the language model can process new fault patterns and configurations through its natural language understanding capabilities, allowing it to dynamically adjust to system changes without retraining while maintaining fast diagnosis speed.
Solution Approach 2:
The language model approach allows for flexible parameter changes in system configurations without requiring model retraining. The model can handle variations in system parameters, configurations, and fault patterns by processing them as natural language inputs, enabling rapid adaptation to changing conditions while maintaining real-time diagnosis capability, unlike traditional models that are sensitive to parameter changes.
4Adaptability or versatility
If language model is used for fault analysis, then adaptability to changing scenarios improves, but computational cost and response time may increase
Solution Approach 1:
The patent applies partial action by using the language model selectively for complex fault scenarios that require adaptability, rather than deploying it for all diagnostic tasks. For routine or simple faults, traditional faster methods can be used, while the language model is invoked only when adaptability is needed. This reduces overall computational energy consumption while maintaining high scenario adaptability when required.
Data Source
AI summary
Embodiments of the present disclosure provide a fault diagnosis method and apparatus for service failure, and a storage medium. The method includes: acquiring abnormal information to be diagnosed described in a natural language; retrieving, from a fault knowledge base, similar target known fault knowledge based on the abnormal information to be diagnosed, where the fault knowledge base is used to maintain a plurality of sets of known fault knowledge; constructing a first fault analysis instruction of the abnormal information to be diagnosed according to the target known fault knowledge and the abnormal information to be diagnosed; and invoking a language model to take the first fault analysis instruction as an input, and output a fault analysis result of the abnormal information to be diagnosed.