SDN Controller Flow Stickiness via BGP-FS Bypass Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data center failure management systems in SDN deployments face challenges in maintaining flow stickiness and scalability, particularly during network failures, which can disrupt service quality and lead to undesirable re-routing of packets.
Innovation Solution
An SDN controller is configured to learn new flows and determine which require flow stickiness, generating commands to redirect packets via bypass routing paths advertised using BGP-FS, avoiding ECMP routes to ensure continuous service delivery during failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ECMP routing is used for load balancing in data centers, then network scalability and resource utilization are improved, but flow stickiness is lost during gateway failures causing service disruption
Solution Approach 1:
The patent segments routing into two distinct paths: ECMP routes for normal load-balanced traffic and BGP-FS bypass routes for sticky flows requiring gateway failover. This segmentation allows simultaneous optimization for both scalability and reliability by directing different traffic types through appropriate paths
Solution Approach 2:
The SDN controller acts as an intermediary that monitors gateway health and dynamically manages routing paths. It inserts BGP-FS route advertisements as a mediator between the control plane and data plane, enabling automatic failover while maintaining flow stickiness without requiring changes to existing ECMP infrastructure
2Reliability
If redundant gateway nodes are deployed for failover, then network reliability is improved, but complex state synchronization and expensive hardware are required
Solution Approach 1:
Instead of maintaining complex state synchronization across redundant gateways, the patent uses BGP-FS to advertise bypass routes that copy the failover capability at the routing layer. This allows gateway redundancy without requiring deep state synchronization, reducing complexity while maintaining reliability
Solution Approach 2:
The patent replaces the mechanical state synchronization mechanism with a software-based BGP-FS routing advertisement system. The SDN controller uses protocol-based route propagation instead of direct state replication, substituting complex hardware/software synchronization with a more manageable routing protocol approach
3Reliability
If BGP-FS bypass routes are advertised for sticky flows, then flow stickiness is maintained during failures, but additional routing overhead is introduced
Solution Approach 1:
The patent applies BGP-FS bypass routing selectively only to flows that require stickiness, rather than all traffic. This partial action approach maintains flow stickiness where needed while avoiding unnecessary routing overhead for flows that can use standard ECMP routing, optimizing the trade-off between reliability and complexity
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A data center failure management system and method in a Software Defined Networking (SDN) deployment. In one embodiment, an SDN controller associated with the data center is configured to learn new flows entering the data center and determine which flows require flow stickiness. Responsive to the determination, the SDN controller generates commands to one or more switching nodes and/or one or more border gateway nodes to redirect the sticky flows arriving at the switching nodes via ECMP routes from the gateway nodes or avoid the ECMP routes by the gateway nodes in order to overcome certain failure conditions encountered in the data center, an external network, or both.