ML-Based Optical Switch Configuration and Deflection Routing in DIDCs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional data center architectures face challenges with latency, scalability, and network management complexity, particularly in handling data-intensive applications, and determining optimal configurations for optical switches and routing paths in Directly Interconnected Data Centers (DIDCs) is a significant challenge.

Innovation Solution

A machine learning-based approach using neural networks to jointly determine optical switch configurations and deflection routing parameters, optimizing interconnect patterns and routing schemes based on traffic matrices, employing backpropagation and gradient descent to minimize latency and ensure efficient network management.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If traditional data center architectures are used, then network management is simpler, but latency increases and scalability deteriorates

Engineering Contradiction:

ImprovelatencyVSAvoidnetwork management complexity

Core Design Contradiction:

SpeedVSDevice complexity

Solution Approach 1:

The patent replaces traditional manual network configuration and routing optimization methods with a machine learning model that automatically learns optimal routing parameters and optical switch configurations from traffic matrices, thereby reducing latency without proportionally increasing management complexity

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system dynamically adjusts routing parameters and optical switch configurations based on learned patterns from traffic matrices, enabling adaptive optimization of network performance metrics such as latency while maintaining scalability in directly interconnected data center architectures

Inventive Principle:

Principle #35Parameter changes

2Productivity

If optimal optical switch configurations are determined manually, then routing paths can be optimized, but the complexity of determining configurations increases

Engineering Contradiction:

Improvenetwork efficiencyVSAvoidconfiguration determination complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The machine learning model performs self-learning from traffic matrices to automatically determine optimal routing parameters and optical switch configurations without requiring manual intervention, thereby improving network efficiency while avoiding the complexity of manual configuration determination

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system uses traffic matrices as feedback input to the machine learning model, which continuously learns and refines routing parameters and optical switch configurations to optimize network performance, eliminating the need for complex manual optimization processes

Inventive Principle:

Principle #23Feedback

3Adaptability or versatility

If directly interconnected data center architecture is implemented, then scalability improves, but determining optimal configurations becomes more difficult

Engineering Contradiction:

ImprovescalabilityVSAvoidconfiguration optimization difficulty

Core Design Contradiction:

Adaptability or versatilityVSDifficulty of detecting and measuring

Solution Approach 1:

The machine learning model performs preliminary learning from traffic matrices to pre-determine optimal routing parameters and optical switch configurations before actual data transmission, enabling directly interconnected data center architecture to achieve scalability without the difficulty of real-time configuration optimization

Inventive Principle:

Principle #10Preliminary action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Enhances network operation with reduced latency, improved scalability, and simplified management by providing optimal optical switch configurations and routing paths in DIDCs.

Implementation Method 1

The machine learning model may include a neural network, and wherein the learning is based at least on backpropagation to learn the first output indicative of the at least one deflection routing parameter and the second output indicative of the at least one optical switch configuration

Methodology Applied

Scientific EffectBackpropagation:

Data Source

PatentUS20260039981A1Machine-learning based approach for robust configuration of optically interconnected data centers

Publication Date: 2026.02.05 NOKIA SOLUTIONS & NETWORKS OY

AI summary

In some embodiments, there may be provided a systems, methods, and articles of manufacture that learn, by the machine learning model and based at least on the at least one traffic matrix, a first output indicative of at least one deflection routing parameter and a second output indicative of at least one optical switch configuration, wherein the learning jointly determines the first output indicative of at least one deflection routing parameter and a second output indicative of at least one optical switch configuration; provide the first output indicative of the at least one deflection routing parameter to a network management system to configure at least one aggregation node comprised in a directly interconnected data center; and provide the second output indicative of the at least one optical switch configuration to the network management system to configure at least one optical switch comprised in the directly interconnected data center.