Network Verification Using Atom-Based Graph Pruning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional network verification methods are inefficient and semi-automated, requiring expertise and failing to detect subtle data plane bugs, while also being unable to map detected bugs to specific controller code, and often miss data plane issues due to incremental verification and lack of real-time detection capabilities.

Innovation Solution

A network verification method that represents forwarding rules using atoms to create a compact, incrementally updated edge-labelled graph, allowing for real-time detection of errors and efficient verification of network-wide reachability properties without raising false alarms, by transforming forwarding behavior representations and pruning redundant meta-information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If conventional network verification methods are used, then expertise is required and detection is semi-automated, but verification efficiency is low and subtle bugs are missed

Engineering Contradiction:
Improveverification automationVSAvoidverification efficiency
Core Design Contradiction:
Extent of automationVSProductivity

Solution Approach 1:

The system performs self-verification by automatically comparing actual forwarding behavior against expected behavior derived from network policies, eliminating the need for manual expert analysis while maintaining high verification efficiency through automated model checking and formal verification techniques

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces manual expert verification processes with automated formal verification systems that use mathematical models and algorithmic approaches to detect subtle bugs, substituting human expertise with computational methods that can systematically analyze network behavior

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Speed

If incremental verification is used, then verification speed is improved, but data plane issues are missed due to lack of real-time detection

Engineering Contradiction:
Improveverification speedVSAvoiddetection accuracy
Core Design Contradiction:
SpeedVSReliability

Solution Approach 1:

The system implements continuous verification by constantly monitoring and comparing actual network forwarding behavior against expected behavior, ensuring that verification is not merely incremental but ongoing and real-time, thus maintaining both speed and reliability through continuous feedback loops

Inventive Principle:
Principle #20Continuity of useful action

3Measurement precision

If detailed forwarding behavior representation is maintained, then verification completeness is improved, but computational overhead increases

Engineering Contradiction:
Improveverification precisionVSAvoidcomputational overhead
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system extracts only the essential forwarding behavior characteristics needed for verification by comparing actual behavior against expected behavior derived from network policies, removing redundant computational elements while maintaining verification precision through focused analysis of critical forwarding paths

Inventive Principle:
Principle #2Taking out (Extraction)

4Difficulty of detecting and measuring

If conventional verification methods are used, then bugs can be detected, but bugs cannot be mapped to specific controller code

Engineering Contradiction:
Improvebug detection capabilityVSAvoidbug location information
Core Design Contradiction:
Difficulty of detecting and measuringVSLoss of information

Solution Approach 1:

The system implements feedback mechanisms that trace detected forwarding behavior deviations back to their source in the controller code and network policies, providing actionable information about where bugs originate by comparing actual behavior against expected behavior and identifying the specific policy or code segment responsible for the anomaly

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS10305776B2Network verification
Publication Date: 2019.05.28 FUJITSU LTD
  • US10305776B2 patent drawing
  • US10305776B2 patent drawing
  • US10305776B2 patent drawing

AI summary

A network verification method may include identifying a data plane for routing traffic through a network that includes one or more forwarding tables for each switch of a plurality of switches in the data plane. The method may also include obtaining a forwarding behavior representation of the data plane based on the forwarding tables, and obtaining a forwarding rule from the one or more forwarding tables. The method may additionally include transforming the forwarding behavior representation based on the forwarding rule. The method may also include pruning meta-information about the forwarding rule that is redundant to forwarding behavior, and comparing the forwarding behavior representation of the data plane to one or more network properties of expected behavior of the data plane to detect one or more errors in the data plane as variations from the expected behavior.