Preempting SLA Violations via QoS Prediction
Solution Overview
Problem
Current SD-WAN implementations are slow to detect and respond to sudden changes in network quality that violate service level agreements (SLAs), leading to delays in rerouting traffic to alternative paths, which can result in continued unsatisfactory performance for application traffic.
Innovation Solution
A device identifies sudden changes in quality of service (QoS) metrics and predicts the duration of these changes, proactively rerouting application traffic to a secondary path to prevent SLA violations using statistical and machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional running mean computation is used to detect SLA violations, then false positives are reduced, but response time to actual SLA violations is delayed
Solution Approach 1:
The system performs preliminary analysis by computing both short-term and long-term running means of QoS metrics. The short-term running mean detects sudden changes quickly, while the long-term running mean provides context to distinguish actual SLA violations from temporary fluctuations. This preliminary dual-timescale analysis enables faster response to genuine violations without increasing false positives.
Solution Approach 2:
The system changes the parameter of measurement timescale by maintaining multiple running mean computations with different window sizes simultaneously. By comparing deviations across these different timescales, the system can quickly identify when a QoS degradation is sustained enough to constitute a real SLA violation, rather than a transient anomaly.
2Measurement precision
If QoS monitoring period is extended to ensure accurate SLA violation detection, then detection accuracy improves, but traffic performance deteriorates due to continued routing over violating paths
Solution Approach 1:
The system takes preliminary action by detecting sudden changes in the short-term running mean and immediately triggering SLA violation declarations. This preliminary detection mechanism allows the system to act before the long-term running mean would normally declare a violation, thereby maintaining traffic performance while still ensuring accurate detection through the multi-timescale verification process.
Solution Approach 2:
The system dynamically adjusts its detection sensitivity by using different running mean window sizes for different detection purposes. The short-term running mean provides dynamic, rapid response to QoS changes, while the long-term running mean provides stable baseline comparison. This dynamic multi-timescale approach enables the system to respond quickly to violations without sacrificing detection accuracy.
Data Source
AI summary
In one embodiment, a device identifies a sudden change in a time series of a quality of service metric for a first path in a network that violates a service level agreement threshold associated with application traffic conveyed via the first path. The device predicts a length of time that the sudden change in the time series will last. The device determines, based in part on the length of time that the sudden change in the time series is predicted to last, that the application traffic should be rerouted onto a second path in the network. The device causes the application traffic to be rerouted onto the second path in the network.