Machine learning approaches for predicting link failures in production networks
Bibliographic record
Abstract
Resolving network failures after they occur through human investigation is a costly and time-consuming process. Predicting upcoming failures could mitigate this to a large extent. In this work, we collect data from a large intercontinental network and study the problem of flapping links, which are indicative of link failures. Such flapping links have their routing metric increased to divert traffic away; this is followed by corrective actions, and eventually their routing metric is lowered again to carry traffic. Using the collected data, primarily metrics reported from Internet Protocol (IP) and optical layers of the network, we develop ML models to predict upcoming link failures. Exploring a sequence of increasingly complex models, we study the relevance of optical metrics, the underlying temporal relations, and the topological relations in improving the predictive model performance. We discovered that optical features such as optical maximum and minimum power or unavailable and errored seconds increased the model’s performance (measured in average precision) by about 9 percentage points while temporal and spatial features improved it further by 8 and 7 percentage points respectively for a total improvement of 24 percentage points. • Optical metrics, in addition to IP metrics, are important when building predictive link failure models. • Capturing time-based relations improves the predictive performance. • Similarly, aggregating data from neighboring links via GNNs is beneficial. • Combining spatial and temporal dimensions, we can develop predictive link failure models with high performance.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.008 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".