Source Routing Optimization via Machine Learning Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In source routing network environments, optimizing the underlying Layer 3 fabric is challenging without tracking and learning source routing behavior over time, as the network fabric lacks information about future paths taken by source routed data flows.
Innovation Solution
Implementing machine learning algorithms that analyze flow information and network characteristics from multiple network elements to predict source routing behavior, providing predictions to source endpoints to optimize source routed network paths and dynamically engineer traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If source routing is implemented without tracking and learning behavior, then source endpoints have routing autonomy, but the network fabric cannot optimize paths
Solution Approach 1:
The system performs preliminary learning and analysis of historical flow information and network characteristics to generate predictions about future routing behavior. This allows the network to proactively optimize paths before actual traffic flows, rather than reacting to observed patterns after the fact.
Solution Approach 2:
The system implements a feedback loop where flow information and network characteristics are continuously collected, analyzed by machine learning models, and used to generate predictions that are sent back to source endpoints. This closed-loop feedback enables continuous optimization of routing paths based on learned patterns.
2Productivity
If machine learning analysis is implemented to predict routing behavior, then path optimization is achieved, but system complexity increases
Solution Approach 1:
The system introduces a machine learning-based prediction engine as an intermediary between raw network data and routing decisions. This intermediary component analyzes flow information and network characteristics to generate actionable predictions, simplifying the overall system architecture by centralizing the complex analysis function.
3Measurement precision
If flow information is collected from multiple network elements, then prediction accuracy improves, but data processing requirements increase
Solution Approach 1:
The system extracts only the most relevant features from the collected flow information and network characteristics, rather than processing all raw data. By identifying and extracting key predictive features, the system maintains high prediction accuracy while reducing the computational burden and energy consumption associated with data processing.
Data Source
AI summary
A controller device sends predictions from a machine learning module to source endpoints. The controller receives flow information and network information from a network elements in a network. The flow information is associated with source routed data flows that traverse the network in source routed network paths. The network information is associated with network characteristics of each of the network elements included in at least one of the source routed network paths. The controller analyzes the flow information and the network information with machine learning to generate a prediction of at least one metric of source routing behavior within the network. The controller sends the prediction of the at least one metric to one or more source endpoints to optimize the source routed network paths used by future source routed data flows originating from the one or more source endpoints.