SDN Switch Rerouting Traffic via Local RTT Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional SDN networks face reduced throughput due to congestion, as the central controller becomes a bottleneck from processing large numbers of probe packets, leading to increased latency and processing burden.
Innovation Solution
Switches in the SDN network collect network statistics directly from the data plane and transmit monitoring packets to determine congestion, allowing for rerouting without relying on the central controller, thereby reducing processing load and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the SDN controller transmits more probe packets at high frequencies to detect network congestion, then the network throughput can be improved through better load balancing, but the processing burden on the controller increases and additional end-to-end latency is introduced
Solution Approach 1:
The patent extracts the congestion detection function from the central SDN controller and relocates it to individual switches. Each switch independently probes its own egress ports by transmitting probe packets and measuring queue depths, removing the burden of processing all probe packets from the controller while maintaining accurate congestion detection capability.
Solution Approach 2:
The patent segments the centralized congestion detection process into distributed operations at each switch. Instead of one controller probing all links, each switch independently probes its own egress ports, dividing the monolithic control function into smaller, parallel units that operate autonomously.
2Productivity
If the SDN controller probes a large number of links at high frequencies to achieve load balancing, then the network throughput is improved, but the controller becomes the bottleneck under heavier traffic rate
Solution Approach 1:
The patent implements self-service by enabling each switch to autonomously detect congestion at its own egress ports without requiring the controller to probe every link. Switches independently transmit probe packets, measure queue depths, and identify congested ports, making the system self-managing at the network edge while the controller focuses on high-level routing decisions.
Solution Approach 2:
The patent extracts the link-probing function from the controller and assigns it to individual switches. This removes the processing burden of handling numerous probe packets from the controller, allowing it to scale better under heavy traffic while maintaining comprehensive network monitoring capability.
3Ease of operation
If traditional network routing algorithm selects the shortest routing path, then the routing is simple and deterministic, but the network throughput is reduced when congestion occurs on the shortest path
Solution Approach 1:
The patent introduces dynamic routing by enabling switches to identify congested egress ports in real-time through probe packet measurement and automatically select alternative egress ports when congestion is detected. This allows routing paths to adapt dynamically to current network conditions while maintaining simplicity through automated decision-making at each switch.
Solution Approach 2:
The patent implements feedback mechanisms where switches continuously monitor queue depths at egress ports using probe packets and use this information to dynamically adjust routing decisions. The feedback loop enables the network to respond to congestion by redirecting traffic away from congested paths, improving throughput while maintaining routing simplicity through automated responses.
Data Source
AI summary
A method for rerouting traffic by a switch in a software defined networking (SDN) arrangement comprises an additional switch, multiple routing paths between the original switch and the additional switch being preconfigured by a SDN controller. The original switch collects network statistics by transmitting monitoring packets on each of the multiple routing paths and then calculating a round-trip time (RTT) value for each path on the return of the monitoring packets to the additional switch, the two switches functioning together as local controllers. After collection of network statistics from all the multiple routing paths, the original switch selects the routing path with the smallest RTT value and reroutes traffic accordingly.