SDN Controller Availability via ML Failure Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Software-defined networking (SDN) controllers are susceptible to failures due to unexpected behavior from virtual network functions and data-plane nodes, leading to instability and performance issues, as traditional dimensioning techniques are insufficient to manage anomalous conditions effectively.

Innovation Solution

A machine learning-based failure prediction system that collects data from SDN controllers and operating environments to develop predictive models, identifying potential failures and triggering corrective actions before they occur, thereby enhancing controller availability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional dimensioning techniques are used to manage SDN controllers, then basic operational stability is maintained, but the system cannot effectively handle anomalous conditions or unexpected behaviors from DPNs and VNFs

Engineering Contradiction:

Improvecontroller availabilityVSAvoidability to handle anomalous conditions

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary actions by collecting status information and usage data before failures occur, training machine learning models on historical data to recognize failure patterns. This allows the controller to predict and prepare for potential failures in advance, rather than merely reacting to them when they happen, thus improving reliability while maintaining adaptability to anomalous conditions.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements continuous feedback loops where status information from the controller and usage information from the operating environment are constantly monitored and fed into machine learning models. The models analyze this feedback to detect anomalies and predict failures, enabling the system to adapt dynamically to changing conditions and maintain high availability even when faced with unexpected behaviors from DPNs and VNFs.

Inventive Principle:

Principle #23Feedback

2Adaptability or versatility

If reactive flow rule installation is implemented to support movable IP addresses and hidden subnets, then network flexibility is improved, but the SDN controller becomes susceptible to failure from unexpected packet flooding

Engineering Contradiction:

Improvesupport for movable IP addressesVSAvoidcontroller stability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The machine learning model performs preliminary analysis of packet patterns and controller status before flooding occurs. By training on historical data including reactive flow rule installation scenarios, the model learns to recognize precursors to packet flooding attacks or anomalous behavior from DPNs and VNFs, allowing the system to take preventive actions before the controller becomes overwhelmed.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system continuously monitors usage information including packet flow patterns, controller response times, and resource utilization. When the machine learning model detects anomalies in this feedback data that precede packet flooding events, it can trigger alerts or automated responses to mitigate the impact on controller stability, thus maintaining reliability while preserving the flexibility of reactive flow rule installation.

Inventive Principle:

Principle #23Feedback

3Loss of information

If the SDN controller processes all packets from DPNs and VNFs, then complete network visibility is achieved, but the controller may be overwhelmed by difficult-to-handle packets

Engineering Contradiction:

Improvenetwork visibilityVSAvoidcontroller processing capacity

Core Design Contradiction:

Loss of informationVSProductivity

Solution Approach 1:

The system implements intelligent feedback mechanisms where the machine learning model analyzes packet patterns, controller load, and status information to dynamically determine which packets require full processing and which can be handled more efficiently or filtered. This allows the controller to maintain necessary network visibility while managing processing capacity through adaptive, intelligence-driven packet handling decisions based on continuous feedback from the system state.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12119981B2Improving software defined networking controller availability using machine learning techniques

Publication Date: 2024.10.15 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)

AI summary

A method of managing a controller of a software defined networking (SDN) network is implemented by a computing device in the SDN network. The method includes receiving status information for the controller, receiving usage information for the operating environment, generating at least one failure prediction for the controller based on the received status information, and outputting prediction information for the at least one failure prediction.