REFORM: Increase alerts value using data driven approach
Bibliographic record
Abstract
We introduce REFORM as a data-driven approach to assess alert quality and increase the value of alerts, reduce alert noise and to identify gaps in alert coverage. In the context of monitoring cloud infrastructure and services, surfacing the right set of alerts to human operators is critical to ensure timely intervention for issues that may otherwise result in significant impact to customers as well as for avoiding operator burnout.To assess alert quality we focus on the notion of actionable alerts – a specific alert is categorized as actionable or not based on historic evidence of operator actions having been taken in response to such alerts. Using alert data collected over a six month period from a combination of cloud environments that include pre-production and production systems, alerts are categorized as true positive, false positive or false negative with respect to their actionability. Based on this, we then introduce and quantify precision and miss rates per alert category (trigger condition) from the historic data. Noisy, non-actionable alerts are identified using precision. Gaps in coverage are identified using miss rate, or failure of monitoring to detect issues.A prescriptive approach is then provided to increase the alerts’ value that includes refinement of alert definitions, replacement of alerts that are no longer actionable and identification of alerting coverage gaps. We monitored our REFORM system over three months and found that the volume of noisy, non-actionable alerts were significantly reduced and furthermore, this resulted in significantly higher intervention rates from the operators.
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 | 0.011 | 0.046 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.004 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.005 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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".