SDN Fabric Failure Prediction via ML Route Disaggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Software defined networking (SDN) fabrics are susceptible to packet loss due to failures, particularly in cloud computing environments where centralized control can lead to vulnerabilities in traffic flows, especially with equal-cost multipath routing (ECMP) which affects network performance and security.
Innovation Solution
A supervisory device uses a machine learning-based failure prediction model to identify affected traffic flows and disaggregates routes for critical subsets to prevent failures, employing a failure modeling engine and flow adjuster to proactively reroute traffic and mitigate the impact of predicted failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If equal-cost multipath routing (ECMP) is used in SDN fabric, then network throughput and load balancing are improved, but susceptibility to packet loss during failures increases
Solution Approach 1:
The system performs preliminary actions by predicting failures before they occur using machine learning models. When a failure is predicted, the system proactively disaggregates routes and reroutes traffic flows to avoid the predicted failure point, thus preventing packet loss before it happens while maintaining the benefits of ECMP during normal operation
Solution Approach 2:
The system applies preliminary anti-action by identifying traffic flows that would be affected by a predicted failure and actively rerouting them away from the potential failure point. This counteracts the harmful effect of packet loss before the actual failure occurs, allowing ECMP to continue providing load balancing while protecting against reliability issues
2Productivity
If centralized control plane decisions are made in SDN, then network performance in cloud environments is improved, but vulnerability to failures in traffic flows increases
Solution Approach 1:
The centralized controller performs preliminary failure prediction using machine learning models and proactively disaggregates routes for affected traffic flows before failures occur. This allows the centralized control plane to maintain performance optimization while preemptively protecting traffic flows from vulnerabilities, ensuring both high performance and reliability
3Reliability
If machine learning-based failure prediction is implemented, then packet loss prevention is improved, but computational complexity and false positives increase
Solution Approach 1:
The system applies partial action by not rerouting all traffic flows when a failure is predicted, but only selecting a subset of critical flows for disaggregation. This reduces computational complexity and minimizes false positive impact while still providing effective packet loss prevention for the most important traffic, balancing reliability improvement with manageable system complexity
Data Source
AI summary
In one embodiment, a supervisory device for a software defined networking (SDN) fabric predicts a failure in the SDN fabric using a machine learning-based failure prediction model. The supervisory device identifies a plurality of traffic flows having associated leaves in the SDN fabric that would be affected by the predicted failure. The supervisory device selects a subset of the identified plurality of traffic flows and their associated leaves. The supervisory device disaggregates routes for the selected subset of traffic flows and their associated leaves, to avoid the predicted failure.