Auriga Intelligent Alert report

  • Release version: Australia
  • Updated May 12, 2026
  • 1 minute to read

    • Auriga Intelligent Alert is an advanced multivariate machine learning (ML) model that learns from historical issues on your instance to provide real-time insight. Auriga monitors your performance metrics to deliver notifications of noteworthy events or deviations from anticipated data patterns.

    Alert time
    Indicates the time when the alert was generated by the Auriga intelligent ML model.
    Alert confidence
    • Low: less than 5% nodes impacted
    • Medium: 5% to 20% nodes impacted
    • High: more than 20% nodes impacted

    An alert confidence score from a model is a numerical value that indicates the model's certainty or confidence in the validity of the alert it has generated. This score typically ranges from 0 to 1 (or 0% to 100%) and helps determine the likelihood that the alert corresponds to a true positive event.

    Nodes affected
    1. From the timestamp when the Auriga model generates an alert, node-level data are retrieved for the following metrics over the preceding 60 minutes:
      • Semaphore Default mean
      • Integration semaphore Qdepth
      • Semaphore Default Qdepth
      • Semaphore AMB send Qdepth
    2. For each of these metrics, identify the node with the maximum value within that hour (there can be multiple nodes with the same maximum value).
    3. After removing duplicates, these nodes will constitute the impacted nodes.
    4. For these impacted nodes, display the maximum value for all the semaphore metrics within the last hour.
    Users session affected
    1. Retrieve the session_summary_loggedin data at the node level for the last hour from the time of the Auriga alert.
    2. Calculate the maximum session_summary_loggedin value for each node within that hour.
    3. For the nodes identified as impacted in the previous step, calculate the sum of the maximum session_summary_loggedin values from the previous step. These will represent the impacted sessions.
    Instance response time
    1. From the time the alert is generated, retrieve the anomaly data for server response time with a 10-minute rollup.
    2. If an anomaly is detected within the past hour, categorize the system as unstable; otherwise, categorize it as stable.