ML-Based SFC Resource Provisioning for Network Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing solutions for provisioning service function chain (SFC) resources rely on static rules and policies, leading to either over-provisioning or under-provisioning, resulting in inefficient resource usage and performance degradation.

Innovation Solution

The implementation of machine learning systems that learn resource consumption tendencies of SFCs to accurately determine the required resources and predict adjustments, optimizing resource allocation through a unified platform combining SDN and NFV technologies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If static rules and policies are used for provisioning SFC resources, then resource provisioning is simplified and easier to manage, but resource usage efficiency deteriorates due to over-provisioning or under-provisioning

Engineering Contradiction:
Improveresource provisioning managementVSAvoidresource usage efficiency
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The patent transitions from static resource provisioning rules to dynamic machine learning-based resource allocation. The system continuously learns from historical resource consumption data and traffic patterns to dynamically adjust resource provisioning decisions, enabling adaptive optimization without manual intervention while improving resource usage efficiency.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent implements a feedback mechanism where the machine learning system continuously monitors actual resource consumption and performance metrics, then uses this feedback to refine future resource provisioning decisions. This closed-loop approach allows the system to learn from past decisions and improve resource allocation accuracy over time.

Inventive Principle:
Principle #23Feedback

2Device complexity

If static rules and policies are used for provisioning SFC resources, then the system complexity is reduced, but resource allocation accuracy deteriorates leading to performance degradation

Engineering Contradiction:
Improveprovisioning system complexityVSAvoidresource allocation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent replaces traditional mechanical rule-based provisioning systems with machine learning algorithms. The ML models process historical data and traffic patterns to make intelligent resource allocation decisions, achieving high accuracy without requiring complex manual rule configurations. The system automatically learns optimal provisioning strategies from data.

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

Solution Approach 2:

The machine learning system performs self-learning and self-optimization by automatically analyzing resource consumption patterns and adjusting provisioning decisions without external intervention. The system serves itself by continuously improving its models based on accumulated data, reducing the need for manual system configuration and maintenance.

Inventive Principle:
Principle #25Self-service

3Productivity

If machine learning systems are implemented for dynamic resource provisioning, then resource usage efficiency is improved through accurate prediction, but system complexity increases

Engineering Contradiction:
Improveresource usage efficiencyVSAvoidprovisioning system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the resource provisioning system into distinct functional modules: data collection components, machine learning model training components, prediction components, and execution components. This modular architecture allows each segment to be developed and optimized independently, managing overall system complexity while enabling advanced ML-based resource allocation.

Inventive Principle:
Principle #1Segmentation

4Measurement precision

If machine learning systems are implemented for dynamic resource provisioning, then resource allocation accuracy is improved, but computational requirements and processing time increase

Engineering Contradiction:
Improveresource allocation accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements preliminary action by pre-training machine learning models offline using historical resource consumption data and traffic patterns. Once trained, the models can make rapid predictions during runtime without requiring extensive real-time computation. This separates the computationally intensive training phase from the time-sensitive prediction phase, achieving both accuracy and speed.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3371696B1System and methods for intelligent service function placement and autoscale based on machine learning
Publication Date: 2021.11.03 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3371696B1 patent drawingFigure 1
  • EP3371696B1 patent drawingFigure 2
  • EP3371696B1 patent drawingFigure 3

AI summary

A method implemented by a computing device to optimize resource usage of service function chains (SFCs) in a network using machine learning. The method includes obtaining, from an autoscale machine learning (ML) system associated with a virtual network function (vNF), a suggested adjustment to an amount of resources provisioned for the vNF. The autoscale ML system is trained online using machine learning to predict an amount of resources to be utilized by the vNF. The autoscale ML system is configured to receive as input an amount of resources currently utilized by the vNF and an amount of resources currently available to the vNF, determine using machine learning the suggested adjustment to the amount of resources provisioned for the vNF based on the input, and output the suggested adjustment. The method further includes providing the suggested adjustment to a resource re-allocator component.