DPI Probe Placement Optimization Using Genetic Algorithms

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The challenge lies in determining the optimal number and positioning of DPI probes in a communication network to minimize costs, ensure efficient flow processing, and prevent network overload, while maintaining quality of service, especially as network size increases.

Innovation Solution

A method utilizing a genetic algorithm to generate and simulate configurations of DPI probes, assessing a cost function that balances probe numbers, bandwidth modification, latency, and unanalyzed streams, iteratively refining solutions until convergence is reached.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If DPI probes are deployed to analyze all flows circulating on the network, then the coverage of DPI analysis is improved, but the financial cost and operational cost increase significantly

Engineering Contradiction:
ImproveDPI analysis coverageVSAvoidnumber of DPI probes
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent applies local quality by deploying DPI probes selectively at specific network nodes rather than uniformly across all nodes. The optimization algorithm determines the precise locations where probes should be positioned to maximize analysis coverage while minimizing the total number of probes deployed. This ensures that each probe is placed in a strategically important location based on network topology and traffic patterns.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements universality by enabling a small number of strategically positioned DPI probes to analyze the majority of network flows. Through intelligent placement optimization, the system achieves comprehensive DPI coverage across the entire network using minimal probe resources, making the probe deployment universally effective throughout the network infrastructure.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Productivity

If DPI probes are positioned to maximize flow analysis, then the DPI analysis efficiency is improved, but the network bandwidth capacity is reduced due to packet redirection

Engineering Contradiction:
ImproveDPI analysis efficiencyVSAvoidavailable bandwidth
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent applies local quality by redirecting packets only to probes located at strategic network nodes rather than routing all packets through centralized analysis points. This localized approach ensures that packet redirection occurs minimally and only where necessary, preserving bandwidth capacity in the rest of the network while achieving effective DPI analysis at specific critical locations.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements preliminary action by pre-optimizing probe positions using network topology analysis and traffic pattern data before actual deployment. This advance planning ensures that probes are positioned to maximize analysis efficiency while minimizing their impact on network bandwidth, avoiding the need for reactive adjustments that could further degrade performance.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If DPI probes are deployed throughout the network, then the flow analysis capability is improved, but the latency time increases due to extended packet paths

Engineering Contradiction:
Improveflow analysis capabilityVSAvoidlatency time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies local quality by positioning DPI probes at network nodes that are strategically close to major traffic flows, minimizing the additional distance packets must travel. This ensures that packet redirection to probes adds minimal latency while still achieving comprehensive flow analysis coverage across the network.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements preliminary action by using network topology analysis and performance data to pre-determine optimal probe locations that minimize latency impact. This advance optimization ensures that probes are placed in positions that maintain short packet paths while still enabling effective DPI analysis, preventing latency issues before they occur.

Inventive Principle:
Principle #10Preliminary action

4Reliability

If the number of DPI probes is increased to cover more flows, then the DPI analysis completeness is improved, but the network overload risk increases

Engineering Contradiction:
ImproveDPI analysis completenessVSAvoidnetwork configuration complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies local quality by deploying probes at specific network nodes with high traffic concentration rather than distributing probes uniformly. This strategic localization allows a small number of probes to capture the majority of network flows, achieving comprehensive analysis coverage while avoiding the complexity of configuring and managing a large distributed probe network.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements universality by designing the probe deployment strategy to achieve maximum analysis coverage with minimal probe resources. The optimization algorithm ensures that each deployed probe serves multiple traffic flows and network functions, reducing the total number of probes needed and thereby simplifying the overall network configuration and management complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP2976855B1Method of determining the number and the positioning of dpi probes in a communication network
Publication Date: 2018.05.16 THALES SA
  • EP2976855B1 patent drawingFigure 1~2
  • EP2976855B1 patent drawingFigure 3
  • EP2976855B1 patent drawingFigure 4

AI summary

This method comprises the steps consisting in: a- on the basis of a group of parent individuals (GPi), generating a population of child individuals (PEi), each individual corresponding to a possible configuration of the DPI component in the network; b- for each individual (Ij) of the population of child individuals (PEi), simulating the behaviour of the network equipped with the DPI component corresponding to said individual; c- evaluating a cost function (H) for said individual; d- selecting a group of child individuals (GEi) exhibiting the lowest values of the cost function; and, while a convergence condition is not complied with, iterating steps a to d by considering as group of parent individuals for the next iteration (GPi+1), the group of child individuals for the current iteration (GEi).