Machine Learning Tuning of Static Network Parameters for Live Traffic

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing network devices, such as CG-NAT, SD-WAN, and flow collectors, operate with static configuration parameters that become inefficient over time due to evolving network conditions, leading to suboptimal performance and user experience.

Innovation Solution

Implementing machine learning to dynamically tune these static parameters in real-time based on monitored patterns in subscriber packet flows, allowing for adaptive adjustments while the devices are operational.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If static configuration parameters are used in network devices, then device complexity is reduced and ease of operation is improved, but network efficiency deteriorates and productivity decreases over time due to evolving network conditions

Engineering Contradiction:
Improvenetwork efficiencyVSAvoidparameter tuning complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system enables self-service by allowing the network device to automatically tune its own configuration parameters using machine learning models. The device monitors its own performance metrics and traffic patterns, then autonomously adjusts parameters without external intervention, resolving the contradiction between maintaining high network efficiency and avoiding complex manual tuning operations.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The invention directly applies parameter changes by dynamically modifying configuration parameters based on machine learning predictions. The system continuously adapts parameters such as buffer sizes, timeout values, and routing thresholds according to real-time network conditions, thereby maintaining optimal network efficiency without requiring complex manual reconfiguration processes.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If static configuration parameters are used, then ease of operation is improved, but adaptability deteriorates as network conditions evolve

Engineering Contradiction:
Improveparameter adaptabilityVSAvoidparameter management ease
Core Design Contradiction:
Adaptability or versatilityVSEase of operation

Solution Approach 1:

The system implements dynamics by transitioning from static to dynamic parameter configuration. Machine learning models continuously predict optimal parameter values based on evolving network conditions, enabling the device to adapt automatically. This dynamic approach maintains high adaptability while preserving ease of operation, as the system handles all adjustments autonomously without requiring user intervention.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The invention applies feedback mechanisms by continuously monitoring network performance metrics and using this information to refine parameter tuning decisions. The machine learning models learn from historical performance data and real-time feedback, enabling the system to adapt to changing network conditions while maintaining simple operation through automated closed-loop control.

Inventive Principle:
Principle #23Feedback

3Productivity

If machine learning is implemented to dynamically tune parameters, then network efficiency is improved, but device complexity increases

Engineering Contradiction:
Improvenetwork performanceVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system uses an intermediary machine learning model that acts as a mediator between network conditions and parameter configuration. Rather than directly managing complex parameter interactions, the ML model processes network metrics and translates them into optimal parameter settings, thereby improving network performance while abstracting away the complexity from the core network device operations.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The invention applies preliminary action by pre-training machine learning models with historical network data before deployment. This pre-processing of knowledge allows the model to make accurate predictions with minimal real-time computation, thereby improving network performance while keeping the runtime system complexity manageable through previously learned patterns rather than complex real-time calculations.

Inventive Principle:
Principle #10Preliminary action

4Adaptability or versatility

If dynamic parameter tuning is implemented, then adaptability is improved, but ease of operation deteriorates due to automated adjustments

Engineering Contradiction:
Improveparameter adaptabilityVSAvoidoperational control
Core Design Contradiction:
Adaptability or versatilityVSEase of operation

Solution Approach 1:

The system maintains ease of operation through self-service automation, where the machine learning model independently handles all parameter tuning decisions without requiring user intervention. The automated adjustments improve adaptability while preserving operational simplicity, as the system manages its own configuration based on real-time network conditions without burdening operators with complex manual tuning tasks.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3806396B1Employing machine learning to predict and dynamically tune static configuration parameters
Publication Date: 2025.10.15 JUNIPER NETWORKS INC
  • EP3806396B1 patent drawingFigure 1
  • EP3806396B1 patent drawingFigure 2
  • EP3806396B1 patent drawingFigure 3

AI summary

A system and method for dynamically altering static parameters on a live network device is disclosed. The system includes a live network device having a plurality of parameters configured thereon that control the application of services to subscriber packet flows and a machine learning device operable to monitor the subscriber packet flows and apply a machine learned model to identify patterns in the monitored subscriber pack flows. The machine learning device is further operable to dynamically alter at least one of the plurality of parameters on the network device based upon the patterns in the monitored subscriber packet flows.