Maintain and monitor

Models are built based on your data. System usage changes over time. Models based on older data might not accurately reflect your current system usage. You can use BMC AMI Ops Insight functionality to evaluate whether your models are still a good match for your data or it's time to generate new models.

BMC AMI Ops Insight’s hybrid‑AI approach ensures that models remain aligned with real workload behavior over time, combining machine learning, AI, and rules‑based analysis with GenAI explanations where enabled.

You can set a schedule to run the model health evaluation and, if required, retrain the models. 

To set the evaluation schedule

1. On the title bar, click the Administration icon . 
2. From the menu on the left of the window, select Model > Model Health Evaluation.
   

3. Select the day and time to run your model evaluation. 
   The data collection starts from the day and time you select. After sufficient data is gathered for evaluation, the calculation is done, and the model evaluation score is displayed in the Model Health chart on the Model Management page. 
   For evaluation, you need at least 15 minutes of data. By default, the evaluation is completed after 60 minutes. For example, if you set the time to 10:00 A.M., the evaluation should be completed by 11:00 A.M. 

   To adjust the value, use the MODEL_HEALTH_DATA_SIZE property in amipdt.properties.
   If BMC AMI Manager is restarted while collecting data for evaluation, the model health score for that evaluation will not be available.

   Success

   Best practice

   For best results, we recommend that you run the evaluation at a time that is close to when the model was generated. Ideally, this would be at peak activity time. Evaluation will not have a noticeable effect on your system's performance.

   Run the evaluation once a week.

4. Click Submit.
   The model evaluation chart is displayed on the Model Management page.
   Model health is also evaluated as part of the model creation process, providing an initial model health score if sufficient data is available.

(Optional) To retrain models

In the Model Health column, you can see the evaluation of the health of your models.

Evaluation is indicated as Poor, Good, and Excellent.

Click on the graph to see the details of the evaluations. If the evaluation indicates a trend, hover over a point for more information.

If you notice that your model's health is trending towards poor, we recommend that you generate a new model. As a general rule of thumb, models are likely to deteriorate after about three months. Results might vary according to the activity levels of your system.

To monitor system health

The BMC AMI Ops Product Health page displays the live status of the product's components and their subcomponents. It indicates the availability of the components and subcomponents. It also displays the number of problems (if any).  
When the Explain Probable Cause feature is available, BMC AMI Assistant can provide plain‑language descriptions of anomaly paths based on BMC AMI Ops Insight’s underlying analysis.

To display the Product Health page, click  at the top right of the navigation panel.

The Product Health page is displayed:

You can click the following options on the top bar or in the tiles to view the current status of the components:

Option	Description
BMC AMI Ops Monitor	Displays the current status of the BMC AMI Ops Monitor server
BMC AMI Ops Insight	Displays the current status of the following subcomponents of the BMC AMI Ops Insight: Docker Container for Workload Graph BMC AMI Manager BMC AMI Manager Database TOMCAT REST interface Docker Container for KPI Graphs Docker Container for PostgreSQL SMF Record Handler Datastream (BMC AMI Datastream for Ops Insight) Db2 Subsystem Scoring Engine
BMC AMI Ops UI Discovery	Displays the current status of the BMC AMI Ops UI Discovery
BMC AMI Ops UI	Displays the current status of the BMC AMI Ops UI server

The subcomponents that are listed below the components are sorted by the severity of the issue, with the most severe at the top.

The health check page has the following options:

Legend	Description
Component	List of components and subcomponents Click to see the subcomponent list.
Info	Displays the component or subcomponent specific information Click Info to see the protocol, host name, job name, port number, and JVM.
Metrics	Displays the metric details such as total memory, free memory, and application CPU usage. Hover over theicon to see the details. Warning Important Currently, not all components display supporting matrices.
Status Message	Displays the statuses of the components The status message provides the details of the status. Warning Important The default value for the following properties in amipdt.properties for USS disk storage is set to: USS_DISK_WARN_THRESHOLD=524288000 USS_DISK_ERROR_THRESHOLD=104857600 If you get one of the following messages in the error status, you must remove irrelevant log files, playbacks, or *.bk* files from the data directory to create free space and prevent the system from running out of space: Storage is below recommended levels. Only n bytes remain. Storage is below minimum levels. Only n bytes remain. For more information, see Managing USS disk storage threshold.
Show me only issues	Display the components that have a problem Select the Show me only issues checkbox to display only those components that have a problem.
Last Seen (GMT)	Displays the latest time and date when a component or subcomponent was last active