Adaptive Network Path Selection via Aggregated Experience Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SD-WAN technologies face challenges in handling fine-grained predictions for SLA violations, leading to an overwhelming number of predictions that network administrators and controllers struggle to manage, and there is a need for longer-term forecasts to optimize user experience.
Innovation Solution
A device predicts the distribution of application experience metrics for multiple paths, aggregates these distributions, and compares them to route traffic through the most favorable subset of paths based on long-term predictions, using a generative model and stateful algorithms to make informed routing decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fine-grained predictions are used for SLA violations, then prediction accuracy and responsiveness are improved, but the number of predictions becomes overwhelming and difficult to manage
Solution Approach 1:
The patent merges multiple fine-grained path-level predictions into aggregated application-level predictions. Instead of managing individual path predictions separately, the system combines them into unified predictions at the application level, reducing the overall number of predictions while preserving accuracy through the aggregation of underlying path metrics.
Solution Approach 2:
The patent creates a universal prediction framework that operates at multiple levels (path level and application level) simultaneously. The same prediction system serves both fine-grained path analysis and coarse-grained application-level decision-making, eliminating the need for separate prediction mechanisms and reducing management complexity.
2Speed
If automatic routing control is enabled based on predictions, then response time to SLA violations is improved, but administrator trust and control are reduced
Solution Approach 1:
The patent introduces prediction aggregations as an intermediary layer between raw path metrics and routing decisions. This intermediary aggregates fine-grained path predictions into application-level insights, providing administrators with summarized information that maintains their control while enabling automated responses based on aggregated trends rather than individual path fluctuations.
Solution Approach 2:
The system performs preliminary aggregation of path predictions into application-level forecasts before triggering routing actions. This preliminary action filters and summarizes the data, allowing administrators to review and approve automated routing decisions based on aggregated application performance trends rather than raw path metrics, thereby maintaining trust and control.
3Duration of action of moving object
If long-term forecasts are used for routing decisions, then user experience optimization is improved, but the granularity and specificity of predictions are reduced
Solution Approach 1:
The patent transitions from a single-dimensional path-level prediction approach to a multi-dimensional framework that operates simultaneously at path level and application level. This dimensional change allows the system to maintain long-term forecast capabilities while preserving granularity through the hierarchical structure that maps application-level trends back to specific path characteristics.
Data Source
AI summary
In one embodiment, a device predicts, for each of a set of paths via which traffic for an online application can be routed, a distribution of an application experience metric for the online application. The device computes, for different subsets of the set of paths, aggregated distributions of their distributions of the application experience metric predicted by the device. The device makes comparisons between the aggregated distributions for the different subsets of the set of paths. The device causes, based on the comparisons, the traffic for the online application to be routed via a particular subset of the set of paths.