Network Failure Mapping via Graph-Based Service Dependency Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As network complexity increases with a large configuration management database (CMDB) containing thousands or millions of items, identifying relationships between items and predicting the effects of a failure becomes challenging, making it difficult to respond effectively to component failures.
Innovation Solution
A method and system that generate a graph of items in a network, where nodes represent items and edges indicate service relationships, allowing for the application of operational rules to determine which items would fail due to the offline of another item, enabling quick identification of affected components and potential re-routing of traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the CMDB contains thousands or millions of items to comprehensively track network components, then the coverage and completeness of network monitoring is improved, but the complexity of identifying relationships between items and predicting failure effects increases
Solution Approach 1:
The patent segments the large-scale network monitoring problem into manageable parts by creating a graph-based representation where items are divided into nodes and relationships are divided into edges. This segmentation allows the system to handle millions of items by processing them as discrete, interconnected units rather than a monolithic structure, thereby maintaining comprehensive monitoring coverage while reducing the complexity of relationship identification.
Solution Approach 2:
The patent introduces a graph data structure as an intermediary between the CMDB and the failure analysis system. This graph serves as a mediator that pre-computes and stores relationship paths between items, allowing rapid failure impact analysis without querying the entire CMDB. The graph intermediary transforms complex relationship queries into simple graph traversal operations, resolving the contradiction between comprehensive monitoring and analysis complexity.
2Measurement precision
If the CMDB stores detailed information about all network items and their relationships, then the accuracy of failure impact prediction is improved, but the time required to identify relationships and respond to failures increases
Solution Approach 1:
The patent applies preliminary action by pre-computing the graph structure that represents all relationships between network items before any failure occurs. The graph is built in advance with nodes representing items and edges representing relationships, so that when a failure occurs, the system can immediately traverse the pre-built graph to identify affected items without performing complex relationship analysis in real-time, thus maintaining high prediction accuracy while minimizing response time.
Solution Approach 2:
The patent creates a copy of the network structure in the form of a graph data structure that mirrors the CMDB relationships. This graph copy contains the essential relationship information needed for failure analysis but in a format optimized for rapid querying. By working with this simplified copy rather than the full CMDB, the system achieves accurate failure impact prediction while significantly reducing the time required to analyze relationships.
3Reliability
If the system analyzes all connected nodes in the graph to determine failure effects, then the completeness of failure impact identification is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies partial action by implementing a depth-limited graph traversal that examines only the necessary portions of the graph to identify failure impacts. Instead of analyzing every single node in the entire graph, the system traverses from the failed node outward to a predetermined depth or until all affected nodes are identified. This approach maintains completeness of failure impact identification for the relevant scope while significantly improving processing efficiency by avoiding unnecessary analysis of unrelated graph portions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is disclosed a method and system for identifying the effects of an item failing in a network. Information about a first item in a network is received. The information indicates that the first item provides a service to other items in the network. A graph of the network is retrieved. A node in the graph that corresponds to the first item is identified. Edges are added to the graph to connect the node of the first item to nodes of the other items that receive the service from the first item. A request is received to model a failure of the first item. The nodes connected to the node that corresponds to the first item are retrieved. Operational rules are applied to each of the retrieved nodes to identify if any of their corresponding items would fail. A list of items that would fail is output.