Predictive Load Balancing for SD-WAN Traffic with Goodput Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current load balancing mechanisms in software-defined wide area networks (SD-WANs) often lead to oscillations between suboptimal routing decisions due to the inability to predictively manage traffic loads across paths with varying bandwidth capacities, resulting in service level agreement (SLA) violations and poor quality of experience (QoE) for applications.
Innovation Solution
A predictive routing process that uses machine learning to collect and analyze network and application telemetry data, predicting path performances for different traffic loads and optimizing traffic balancing across multiple paths to prevent SLA violations and enhance QoE by proactively rerouting traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If load balancing mechanisms spread full traffic load across multiple paths, then traffic distribution is improved, but routing stability deteriorates due to oscillations between suboptimal decisions
Solution Approach 1:
The system performs preliminary actions by predicting future path performances and SLA violations before they occur. The machine learning model analyzes current and historical path characteristics to forecast future conditions, enabling proactive traffic engineering decisions that prevent oscillations rather than reacting to them after they occur.
Solution Approach 2:
The load balancing mechanism dynamically adjusts traffic loads across paths based on predicted future performances rather than static or reactive adjustments. The system continuously updates traffic load allocations as predictions change, creating a dynamic response that adapts to evolving network conditions while maintaining stability.
2Reliability
If machine learning-based routing predicts and reroutes traffic in advance, then SLA compliance is improved, but system complexity increases
Solution Approach 1:
The system introduces an intermediary machine learning model that sits between raw path performance data and routing decisions. This intermediary processes and interprets complex network telemetry data, translating it into actionable predictions about future SLA compliance, thereby simplifying the decision-making process while maintaining high reliability.
Solution Approach 2:
The patent replaces traditional mechanical or rule-based routing decision systems with a machine learning-based predictive system. Instead of relying on static routing tables or simple threshold-based reactions, the system uses ML models to predict future path performances and make informed routing decisions, substituting complex mechanical control with intelligent algorithms.
3Ease of operation
If traffic load is dynamically adjusted across paths, then QoE is improved, but measurement and prediction difficulty increases
Solution Approach 1:
The system implements feedback mechanisms where actual path performances and QoE metrics are continuously measured and fed back to the machine learning model. This feedback loop allows the model to learn from real-world outcomes and improve its predictions, making the complex task of measuring and predicting path performance more manageable through iterative learning.
Solution Approach 2:
The system changes parameters by using multiple different metrics and features for predicting path performance, such as historical delay, jitter, packet loss, and traffic patterns. By analyzing multiple parameters simultaneously and changing their weights based on learned importance, the system simplifies the overall prediction task while improving QoE measurement accuracy.
Data Source
AI summary
In one embodiment, a device obtains data indicative of quality of experience for an online application. The device predicts, based on the data, path performances of network paths between an endpoint and the online application for different traffic loads. The device selects traffic loads for the network paths between the endpoint and the online application, based on the path performances predicted by the device. The device causes application traffic to be load balanced across the network paths between the endpoint and the online application, in accordance with those traffic loads selected by the device.