All Products
Search
Document Center

Database Autonomy Service:Performance anomaly detection

Last Updated:Jun 20, 2026

Database Autonomy Service (DAS) uses machine learning and intelligent algorithms to detect and predict anomalies in core database instance metrics. It integrates with the one-click diagnostics module to pinpoint the root cause of an anomaly.

Prerequisites

The database instance must meet the following requirements:

Procedure

  1. Log on to the DAS console.

  2. In the navigation pane on the left, click Intelligent O&M Center > Instance Monitoring .

  3. Find the target instance and click the instance ID to open the instance details page.

  4. In the left-side navigation pane, click Dashboard > Exception Detection.

  5. On the Exception Detection page, select a time range to view anomaly detection and prediction results for the database metrics.

    Note

    When you select a time range, the end time must be later than the start time. The time range cannot exceed seven days.

    • Click More Metrics to select the metrics for anomaly detection and prediction. For descriptions of the metrics, see Monitoring metrics.

    • In the Exception Information area, click Diagnose for an abnormal metric to open the Diagnostic Tree page. On this page, you can view the overall performance of the database instance at the time of the exception and quickly identify the cause. The Exception Information area displays the detected abnormal metrics, such as mysql.qps and mysql.mem_usage, along with their exception times and abnormal values. Click the Diagnose link for a target metric to open the diagnosis details page.

Monitoring metrics

DAS currently supports performance anomaly detection for the following metrics.

Metric

Description

tps

The number of transactions per second.

qps

The number of queries per second.

active_session

The number of active sessions.

delete_ps

The average number of DELETE statements executed per second.

insert_ps

The average number of INSERT statements executed per second.

update_ps

The average number of UPDATE statements executed per second.

select_ps

The average number of SELECT statements executed per second.

bytes_received

The average number of bytes received from all clients per second.

bytes_sent

The average number of bytes sent to all clients per second.

innodb_bp_hit

The read cache hit rate of the InnoDB buffer pool.

innodb_data_written

The average number of bytes written by InnoDB per second.

innodb_data_read

The average number of bytes read by InnoDB per second.

mysql.innodb_log_writes

The average number of physical writes to the InnoDB redo log file per second.

innodb_rows_deleted

The average number of rows deleted from InnoDB tables per second.

innodb_rows_read

The average number of rows read from InnoDB tables per second.

innodb_rows_inserted

The average number of rows inserted into InnoDB tables per second.

innodb_rows_updated

The average number of rows updated in InnoDB tables per second.

mysql.mem_usage

The memory usage of the MySQL instance as a percentage of the total operating system memory.

mysql.cpu_usage

The CPU utilization of the MySQL service process, as a percentage. The maximum value for Alibaba Cloud database instances is 100%.