All Products
Search

This Product

    Database Autonomy Service:Anomaly detection

    Document Center

    Database Autonomy Service:Anomaly detection

    Last Updated:Jul 17, 2024

    Anomaly detection is a major consideration in routine O&M of databases. Database Autonomy Service (DAS) provides the anomaly detection feature to detect exceptions 24/7 based on machine learning and fine-grained monitoring data. This detection mechanism allows DAS to detect database exceptions faster than traditional threshold-based alerting mechanisms. This topic describes the benefits of the anomaly detection feature and how to view the detection results.

    Prerequisites

    • The database instance that you want to manage is of one of the types described in the following table.

      Database

      Region

      • ApsaraDB RDS for MySQL

      • ApsaraDB MyBase for MySQL

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Heyuan), China (Guangzhou), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Ulanqab), China (Nanjing - Local Region), China (Fuzhou - Local Region), China (Chengdu), China (Zhengzhou - Local Region), China (Hong Kong), Japan (Tokyo), South Korea (Seoul), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Philippines (Manila), Thailand (Bangkok), Australia (Sydney), India (Mumbai) Closing Down, UAE (Dubai), SAU (Riyadh - Partner Region), Germany (Frankfurt), US (Silicon Valley), US (Virginia), and UK (London)

      ApsaraDB RDS for PostgreSQL

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Heyuan), China (Guangzhou), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Ulanqab), China (Chengdu), China (Hong Kong), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Australia (Sydney), India (Mumbai) Closing Down, UAE (Dubai), Germany (Frankfurt), US (Silicon Valley), US (Virginia), and UK (London)

      ApsaraDB RDS for SQL Server

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Heyuan), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Ulanqab), China (Chengdu), China (Hong Kong), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Australia (Sydney), India (Mumbai) Closing Down, UAE (Dubai), Germany (Frankfurt), US (Silicon Valley), US (Virginia), and UK (London)

      PolarDB for MySQL Standard Edition and Cluster Edition

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Heyuan), China (Guangzhou), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Ulanqab), China (Chengdu), China (Hong Kong), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Philippines (Manila), Australia (Sydney), India (Mumbai) Closing Down, Germany (Frankfurt), US (Silicon Valley), US (Virginia), and UK (London)

      ApsaraDB for Redis

      • Community Edition instances

      • Enhanced Edition (Tair) DRAM-based instances

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Heyuan), China (Guangzhou), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Chengdu), China (Hong Kong), Japan (Tokyo), South Korea (Seoul), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Philippines (Manila), Thailand (Bangkok), Australia (Sydney), India (Mumbai) Closing Down, UAE (Dubai), SAU (Riyadh - Partner Region), Germany (Frankfurt), US (Silicon Valley), US (Virginia), and UK (London)

      Enhanced Edition (Tair) persistent memory-optimized and ESSD/SSD-based instances

      China (Hangzhou), China (Shanghai), China (Shenzhen), China (Beijing), China (Zhangjiakou), China (Hong Kong), Singapore, Germany (Frankfurt), and US (Virginia)

    • The database instance is connected to DAS and is in the Normal Access state.

      Note

      For more information about how to connect a database instance to DAS, see Connect an Alibaba Cloud database instance to DAS.

    Benefits

    The anomaly detection feature detects exceptions 24/7 based on machine learning and fine-grained monitoring data. Compared with rule-based or threshold-based alerting, you can detect database exceptions in a more timely manner.

    Item

    Traditional solution

    DAS anomaly detection

    Method

    Rule-based or threshold-based

    AI-based

    Monitored objects

    Metrics

    A wide range of objects, such as metrics, SQL statements, logs, locks, and O&M events

    Latency

    From 5 minutes to one or more days

    Quasi-real-time

    Detection method

    Fault-driven

    Exception-driven

    Periodic detection

    Not supported

    Automatic and periodic

    Adaptability

    Not supported

    Adaptive to services that have different characteristics

    Prediction

    Not supported

    Supported

    View anomaly detection results

    In the autonomy center, you can view events that are detected within a specific time range.

    1. Log on to the DAS console.

    2. In the left-side navigation pane, click Instance Monitoring.

    3. On the page that appears, find the database instance that you want to manage and click the instance ID. The instance details page appears.

    4. On the instance details page, click Autonomy Center in the left-side navigation pane.

    5. Specify a time range to view the exception detection results within the time range.

    Enable event subscription

    After you enable the event subscription feature for a database instance, DAS sends you a notification every time a subscribed event is triggered. You can specify a notification method such as text messages based on your business requirements. For more information, see Event subscription.

    Note

    To receive notifications about anomaly events, set the urgency level of the events to Warning. You can specify an urgency level based on your business requirements.

    FAQ

    Q: How is the change rate of the related metrics in the Analysis of Abnormal Metrics section of the Anomaly Snapshots tab of the Anomaly Detection of Metrics (Time Series Anomaly Detection) event calculated?异常指标

    A: Change rate of a metric = Actual metric value/Predicated metric value. DAS uses the data at a granularity of hours of a database instance in a specific period of time to predict the metric value of the database instance in the current time range. The predicted metric value is used as a baseline and compared with the actual metric value. This way, the change rate of the metric is calculated.

    References

    You can use the autonomy features of DAS to allow DAS to automatically handle database exceptions. For more information, see the following topics: