Data Plane Route Failover in SDN Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud networking systems face significant delays in route updates due to the propagation of changes across a large number of components, which can exceed service level objectives and impact reliability, especially when using protocols like BGP and BFD, leading to high latency in failover and packet loss mitigation.
Innovation Solution
Implementing a method that maintains primary and secondary routes within the data plane, where the data plane components monitor their own health and automatically switch to a secondary route upon failure, using Border Gateway Protocol (BGP) and Bidirectional Forwarding Detection (BFD) protocols to ensure fast and reliable failover routing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If BGP protocol is used to propagate route updates across cloud network components, then routing information can be distributed to all endpoints, but propagation delays increase significantly (several minutes) which exceeds service level objectives
Solution Approach 1:
The patent segments the monolithic BGP propagation process into distributed route set management at each data plane component. Instead of centralized BGP propagation to all endpoints, each component independently maintains route sets with primary and secondary routes, enabling localized decision-making and eliminating the need for widespread control plane updates.
Solution Approach 2:
The patent implements preliminary action by pre-configuring secondary routes alongside primary routes in the data plane before failures occur. When a primary route fails, the secondary route is immediately activated without requiring control plane intervention or propagation delays, thus achieving fast failover within service level objectives.
2Measurement precision
If BFD protocol is used to detect link faults between routers, then failure detection capability is improved, but dependency on control plane propagation mechanisms still causes sub-second failover requirements to be unmet
Solution Approach 1:
The patent extracts the failover decision-making process from the control plane (BGP/BFD) and places it directly in the data plane. Data plane components monitor their own health status and independently execute failover to secondary routes without waiting for control plane detection and propagation, thereby eliminating the bottleneck that prevents sub-second failover.
Solution Approach 2:
The patent implements self-service by enabling data plane components to autonomously monitor their own operational status and automatically switch to secondary routes when failures are detected. This self-managed approach eliminates dependency on external control plane mechanisms for failover execution, achieving fast response times.
3Reliability
If configuration changes are distributed to all endpoints in software-defined networks, then routing updates can be propagated, but convergence time cannot be bounded and may exceed service level objectives
Solution Approach 1:
The patent segments the routing convergence process by maintaining route sets locally at each data plane component rather than distributing configuration changes across all endpoints. This segmentation allows independent, parallel convergence at each component without coordination overhead, bounding convergence time within service level objectives.
Solution Approach 2:
The patent applies preliminary action by pre-establishing secondary routes in the data plane before failures occur. When primary routes fail, the pre-configured secondary routes enable immediate failover without requiring time-consuming configuration distribution or convergence negotiations across the network.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosed technology consolidates the switch over of dynamic routes to a centralized location, wherein the dynamic routes are tied to the data plane itself. Detection of a health failure within a primary route allows the cloud network and the associated virtual network stack to transfer packet routing to a pre-programmed or configured secondary route.