Distributed Message Routing Using Recursive Least Squares Latency Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network routing techniques, such as Shortest Path First, face performance issues due to race conditions and congestion when networks are under high load, leading to oscillations that degrade network performance in virtualized Message-Oriented Middleware (MOM) systems.
Innovation Solution
A distributed scheduling procedure using a recursive least-squares prediction model to estimate message latency, combined with choke packets for congestion control, allows message routers to select nodes that minimize latency without introducing a single point of failure, ensuring efficient and scalable message routing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If Shortest Path First routing is used, then routing simplicity is improved, but network performance degrades under high load due to race conditions and congestion
Solution Approach 1:
The patent applies dynamics by transitioning from static shortest path routing to dynamic latency-based routing. The system continuously monitors actual message latency through choke packets and updates routing decisions in real-time, allowing the network to adapt to changing load conditions and avoid race conditions that degrade performance under high load.
Solution Approach 2:
The patent implements feedback mechanisms through choke packets that carry latency information from destination nodes back to source nodes. This feedback loop enables routing agents to learn actual message latency experience and adjust future routing decisions accordingly, resolving the performance degradation issue by incorporating real network conditions into routing choices.
2Loss of time
If latency-based routing is implemented, then message latency is reduced, but system complexity increases due to prediction models and feedback mechanisms
Solution Approach 1:
The patent applies preliminary action through the use of prediction models that estimate future message latency before routing decisions are made. By predicting latency trends based on historical data and choke packet information, the system proactively selects optimal routes rather than reacting to congestion after it occurs, reducing message latency while keeping the prediction mechanism relatively simple.
Solution Approach 2:
The patent uses copying by having multiple nodes replicate the same routing logic and prediction models. Each node maintains its own simplified prediction mechanism and routing agent, allowing the system to distribute complexity across multiple identical or similar components rather than requiring a single complex centralized controller.
3Productivity
If multiple nodes are added to virtualized ESB, then message processing bandwidth is improved, but routing decision complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the routing decision process into independent segments at each node. Each node's routing agent operates autonomously using local latency predictions and choke packet information, making routing decisions independently rather than requiring complex centralized coordination. This segmentation allows the system to scale bandwidth by adding nodes without proportionally increasing routing decision complexity.
Solution Approach 2:
The patent uses parameter changes by focusing routing decisions on a single key parameter - predicted message latency - rather than considering multiple complex factors simultaneously. By simplifying the routing criterion to latency minimization based on prediction models and feedback, the system can handle increased numbers of nodes without experiencing proportional increases in routing decision complexity.
Data Source
AI summary
Routing data packet traffic in packet networks, expediting the flow of the data packets and reducing the effects of congestion at nodes is accomplished by capturing latency data for nodes for initial measurement times; and computing a least squares estimate of the delay as a recursive least-squares solution.