Dependency Graph Incident Linking for Cloud Services
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing incident linking models in large-scale cloud services struggle to accurately and efficiently manage incidents due to interdependencies between services, leading to alert storms and increased manual toil for On-call engineers (OCEs).
Innovation Solution
The system generates and updates a dependency graph that aligns with textual information about incidents, using a machine learning model to perform incident linking by combining graph data and text data. This approach improves the accuracy of suggested links between incidents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing incident linking models use only textual and contextual information, then the system is simpler to implement, but the accuracy of incident linking deteriorates when incidents arise from interdependent services
Solution Approach 1:
The patent merges graph data representing service dependencies with textual incident information to create a hybrid linking approach. The system combines dependency graph structures with text embeddings to achieve more accurate incident linking, particularly for cascading failures in microservices architectures where traditional text-only methods fail.
Solution Approach 2:
The patent introduces an intermediary alignment mechanism that projects text embeddings into the graph embedding space. This alignment layer acts as a mediator between textual incident descriptions and graph-based service dependencies, enabling the system to leverage both data types effectively for accurate incident linking.
2Reliability
If the system processes all incident data to improve linking accuracy, then the reliability of incident management improves, but the processing time and computational resources increase
Solution Approach 1:
The patent segments the incident data processing into distinct components: extracting textual features, retrieving relevant dependency graph subgraphs, aligning embeddings, and generating links. This segmentation allows the system to process only relevant portions of the graph data (subgraphs) rather than the entire dependency graph, reducing computational overhead while maintaining reliability.
Solution Approach 2:
The system performs partial action by retrieving and processing only the subgraph portions relevant to each incident rather than processing the complete dependency graph. This selective processing approach maintains high reliability for incident linking while significantly reducing processing time and computational resources required.
3Measurement precision
If the dependency graph includes all historical correlating links, then the accuracy of incident linking improves, but the graph becomes more complex and harder to maintain
Solution Approach 1:
The patent implements a dynamic dependency graph that is periodically updated with new correlating links from incident data. The graph structure adapts over time, incorporating historical relationships that improve linking accuracy while maintaining flexibility to reflect current service dependencies. This dynamic approach balances accuracy with manageability.
Data Source
AI summary
Systems and methods are provided for generating and updating a dependency graph that is used in combination with textual information about incidents to improve incident-linking suggestions. Systems and methods are also provided for generating, training, and using a machine learning model configured to perform incident linking using both graph data and text data. Beneficially, these systems and methods align the graph data and text data in order to more efficiently and accurately leverage information from the multi-modal data.