Overlay Network Routing via Underlay Representation Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network technologies face challenges in optimizing traffic routing on overlay networks without knowledge of the underlying opaque underlay network's topology and routing, leading to inefficiencies in traffic management and performance.
Innovation Solution
A machine learning model is trained using traffic matrices and measured delays to learn a representation of the underlay network's routing and background traffic, allowing for the optimization of overlay network routing by minimizing the difference between average and measured delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional routing optimization methods are used on overlay networks without underlay network knowledge, then routing decisions can be made independently, but network performance optimization is limited due to lack of visibility into actual traffic conditions and topology
Solution Approach 1:
The patent creates a virtual copy of the underlay network within the overlay network controller. This copy includes topology information, routing policies, and traffic state data that are replicated from the underlay network. By maintaining this virtual representation, the overlay network can make informed routing decisions without direct access to the physical underlay infrastructure, thus resolving the information asymmetry while preserving independent control.
Solution Approach 2:
The patent introduces an intermediary component (the network controller with machine learning model) that acts as a bridge between the overlay and underlay networks. This intermediary learns and models the underlay network characteristics, then uses this knowledge to optimize overlay routing. The intermediary translates underlay network conditions into actionable routing information for the overlay, enabling performance optimization without direct underlay access.
2Measurement precision
If machine learning models are trained with complete underlay network information, then routing optimization accuracy improves, but system complexity and data requirements increase
Solution Approach 1:
The patent applies partial action by training the machine learning model with a carefully selected subset of underlay network information that is most relevant for routing optimization. Rather than incorporating all possible underlay data (which would increase complexity), the model focuses on key parameters such as topology structure, routing policies, and aggregate traffic state. This selective approach achieves sufficient measurement precision while controlling model complexity.
Solution Approach 2:
The patent transforms complex underlay network information into simplified parameters suitable for machine learning processing. Raw network data is converted into meaningful features such as delay measurements, traffic matrices, and topology representations. This parameter transformation reduces data dimensionality and complexity while preserving the essential information needed for accurate delay prediction and routing optimization.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
In some embodiments, there may be provided a method that includes receiving, as a first input to a first machine learning model, at least a first traffic matrix indicative of an amount of traffic routed among at least one node pair of an overlay network; receiving, as a second input to the first machine learning model, information indicative of overlay network routing among the at least one node pair of the overlay network; receiving, as a third input to the first machine learning model, measured delay between the at least one node pair of the overlay network; and learning, by the first machine learning model, a representation of an underlay network, the learning using a minimization of a difference between an average delay in the underlay network and the measured delay between the at least one node pair of the overlay network.