Expand Your Alert Coverage With Recommended Conditions
Blog post from New Relic
Just as smoke detectors are essential for alerting homeowners to fires, production applications require alerts to signal when issues arise, though setting precise alert conditions can be challenging. New Relic One offers an Alert Condition Recommendation service that employs AI and machine learning to suggest specific metrics and signals for monitoring, allowing users to customize alerts according to their needs. The service leverages entity tags to enhance the precision of recommendations, dividing entities into clusters based on tag similarity and identifying common conditions for each cluster. In cases where data is insufficient, the model refers to a community-curated golden signals data set. The system also suggests appropriate thresholds for alerts by analyzing historical data, aiming to balance between over-alerting and missing critical issues. This service is designed to improve alert coverage efficiently, ensuring comprehensive system monitoring akin to having functional smoke detectors at home.