Multi-domain Network Path Computation via Matrix Fragments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-domain network path computation methods fail to meet the requirements of minimal model and data sharing between domains, particularly in centralized and distributed approaches, which are either too invasive or limited in concurrency and path exploration.
Innovation Solution
The proposed solution involves a diffusive tracing approach with a Query Orchestration Layer that concurrently forms multiple matrix fragments across domains, allowing for minimal model reconciliation and data sharing, and short-circuits fruitless work by broadcasting the lowest weight, enabling concurrent and asynchronous path computation with exhaustiveness and observability of completion conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a centralized approach is used for multi-domain path computation, then path computation capability is improved, but model sharing and data sharing between domains increases
Solution Approach 1:
The patent segments the centralized path computation problem into multiple independent matrix fragment computations, where each domain computes its own local matrix fragment based on its internal topology and constraints. This segmentation allows domains to maintain privacy while contributing to the global path computation solution.
Solution Approach 2:
The patent introduces an intermediary mechanism (the matrix fragment exchange protocol) that enables domains to share only essential routing information without exposing internal network details. The matrix fragments act as intermediaries that carry necessary path computation data while preserving domain privacy.
2Loss of information
If distributed router/protocol-based approach is used, then data sharing between domains is minimized, but constraint support is limited
Solution Approach 1:
The patent changes the parameter representation from detailed network state information to compact matrix fragments that encode path computation constraints. This parameter transformation allows distributed computation to support arbitrary NMS constraints while minimizing data sharing, as constraints are expressed in a standardized matrix format.
3Device complexity
If conventional crankback approaches are used, then implementation simplicity is maintained, but concurrency and path exploration are limited
Solution Approach 1:
The patent applies preliminary action by pre-computing matrix fragments for multiple potential paths before actual path computation is needed. This allows the system to explore multiple paths concurrently and select the optimal one, rather than sequentially trying paths with crankback.
Solution Approach 2:
The patent maintains continuity of useful action by enabling parallel computation of multiple path candidates simultaneously through the matrix fragment approach. Unlike crankback which pauses and retries, this approach continuously computes and evaluates multiple paths in parallel.
Data Source
AI summary
An orchestration layer includes one or more processing devices communicatively coupled to a plurality of domains in a multi-domain network, wherein the one or more processing devices are configured to receive a request for a path, wherein the path is requested from a source in a first domain to a destination in a second domain in a multi-domain network including a plurality of domains, relay the request to each domain in the multi-domain network, and wherein each domain in the multi-domain network is configured to compute a matrix fragment and the matrix fragments from a plurality of domains are used to find the path, and provide a response with the path from the source in the first domain to the destination in the second domain.