Managing projection models



BMC AMI Ops Insight compares models, which are based on normal behavior in your environment, to current behavior to see if you are experiencing problems. You can have models for LPARs and for subsystems. You can also combine subsystem models that belong to the same data sharing group to create sharing group models.

To access the Manage Models page, click the Gear Icon.PNG icon in the title bar. The page includes three tabs:

  • Current – Shows models that are currently available for use
  • Pipeline – Shows models that are currently being built
  • Archive – Shows models that are no longer available but are kept for reference

Use the following procedures to manage models in BMC AMI Ops Insight:

All actions, except for creating a model, cloning, and merging, support multiple selection. 

To create new Subsystem/LPAR models

The product creates new models by analyzing data from your system in a process called training. Use this procedure to set up a new training request.

  1. Click the Administration icon Gear Icon.PNG in the title bar. 
  2. On the Manage Models page, click New Model.
  3. Click Subsystem_LPAR_button.png.
    Create_New_Model.png
  4. In General Information, enter the following settings:

    Field

    Description

    Data Set Name

    Name of the historical SMF data set that contains the source data

    The data set must follow TSO data set naming conventions. Tape data sets and migrated data sets are not supported. 

    You can use up to 50 data sets for a single model. Enter each data set on a separate line.

    Sort SMF

    Select this to use BMC's sorting algorithm (recommended) to sort the model data by date and time, to minimize resource consumption. Unless you have a specific issue that prevents you from using BMC's sorting algorithm, leave this selected.

    Important

    If you select Sort SMF, Pacing Record Count is disabled.

    Pacing Record Count

    Number of records processed before there is a pause in the training process
    Use Pacing Record Count to control resource usage of your training request. When the number of records is processed, operations are suspended for an interval equivalent to the amount of time it took to process the previous batch of records and to write them to the output file.

    Default: By default pacing is turned off.

    Important

    If you select Sort SMF, Pacing Record Count is disabled.

    Date Range of Model Data

    Start date/time and end date/time for the historical SMF data used in the model

    The date range can be smaller than the full date range covered by the data in the data set.

    Dates and times are according to the local date/time of the data.

    Best practice
    We recommend the following for creating models:

    • For models based on SMF 100 records, use at least a month of data.
    • For models based on SMF 70-79 records, use at least 8 weeks of data.


    Make sure that the data you use does not include any abnormal behavior. If the data you use to create your model includes abnormal behavior, it skews the model and the product cannot provide accurate projections.

    Label / Short Description

    Description for the model This description is displayed on the list of models to help you identify your model.You can use this as an alias for the model name to allow you to easily identify the model and to differentiate it from your other models.

    Folder Name

    USS subdirectory path to store the model bundles or playback file

    The product searches for this folder within the DAT folder in your product installation. If the folder does not exist, the product creates it.

    The path must follow USS naming conventions. You must have write access for the folder and sufficient space to store the data.

  5. Click Next_button.png.
  6. In  Model Type & Details , select the Model Type and enter the details:
    New_Model_Type.png

    Important

    You can create models of both Model Types in the same training request.

    Model Type

    Field

    Description

    LPAR

    Sysplex

    Creates models for all LPARs in the SysplexSysplex can be up to 8 characters long and must match the SMF70XNM field in the SMF data.


    LPAR

    Creates a model only for the individual LPARLPAR can be up to 8 characters long and must match the SMF70SNM field in the SMF data.

    Db2 Subsystem

    Data Sharing Group

    Creates models for all subsystems in the Data Sharing Group

    Data Sharing Group name can be up to 8 characters long and must match the QWHADSGN field in the SMF data.


    Db2 Subsystem Name

    Creates a model only for the individual subsystemDb2 Subsystem Name can be up to 4 characters long and must match the QWHAMEMN field in the SMF data.

  7. Click Submit_Button.png.
    The new model is added to the Pipeline tab on the Manage Projection Models page. Its status is In Progress until the training process is complete. Depending on the amount of data in the training request and the pacing settings, this may take a very long time.  
    • If the training process encounters an error, the status changes to  Error . Click on the Error status to see details about the error.
    • If you want to cancel the training while it is in progress, click Cancel Request.
    • If you have a training request in progress, you cannot issue another training request. If you do, an error message is displayed.

When the training process is complete, the new model moves to the Current tab. If you selected Db2 Sharing Group, a separate model entry is added for each subsystem in the sharing group. If you selected Sysplex, a separate model entry is added for each LPAR in the Sysplex.

Click on the name of a model to see the details for the model.

To create a new sharing group model

Use this procedure to merge subsystem models into a sharing group model.

  1. Click the Administration icon  Gear Icon.PNG in the title bar. 
  2. On the Manage Models page, click New Model.
  3. Click Sharing_group_button.png.
  4. On the Sharing Group tab, select a sharing group.
  5. On the Date Range tab, select a date range.

    Important

    To successfully create a sharing group model, you must select a date range that is included in all of subsystem models that you include.

  6. On the Subsystem Models tab, select subsystem models that you want to merge.
  7. On the Details tab, enter a valid Output Path and a Label for the sharing group model.
  8. Click merge_button.png.

To clone a model

Use this procedure to create a new model based on an existing model.

  1. Click the Administration icon Gear Icon.PNG in the title bar. 
  2. Click the Actions menu for the model you want to clone and select Clone model request.
    The Create New Model dialog opens with the fields populated with entries from the original model.
  3. Modify the fields as necessary and click Submit_Button.png.

To select models

Use this procedure to select a model for BMC AMI Ops Insight to use for scoring, to compare with your system's current behavior.

  1. Click the Administration icon Gear Icon.PNG in the title bar. 
  2. Do one of the following: 
    • To select a single model, click the Actions menu.
    • To select multiple models, for each model select the checkbox next to the Model Name and click the Multiple_Actions_button.pngmenu at the top of the list.

      Important

      Loading multiple models simultaneously may take a long time.

  3. Do one of the following:
    • To select the models immediately, select  Load Model (Select) .
      The Status of models that are selected changes from Available status.PNG to Selected status.PNG

      Important

      Only one model can be selected for each subsystem at any one time.

    • To select the models the next time BMC AMI Ops Insight is restarted, select Load (Select) for restart.

      Important

      Models that are set to load on restart are indicated by a star icon. If you want to cancel the load on restart, select Disable - Load for restart from the Actions menu.

Models that are currently being used for scoring are marked with a lightning icon Lightning_Icon.png.

To deselect models

Use this procedure to deselect a model that BMC AMI Ops Insight has been using.

Important

You cannot deselect models that are currently being used for scoring.

  1. Click the Administration icon Gear Icon.PNG in the title bar. 
  2. Do one of the following: 
    • To deselect a single model, click the Actions menu.
    • To deselect multiple models, for each model you want to deselect, select the checkbox next to the Model Name and click the Multiple_Actions_button.pngmenu at the top of the list:
  3. Select Unload model (De-select).

The Status of models that are deselected changes from Selected status.PNGtoAvailable status.PNG.

To archive models

If you don't want to use a model, but you want to keep it for later reference or later use, you can move it to the archive.

Important

You can only archive models that are in Available status.PNG status.

  1. Click the Administration icon Gear Icon.PNG in the title bar. 
  2. Do one of the following: 
    • To archive a single model, click the Actions menu.
    • To archive multiple models, for each model you want to archive, select the checkbox next to the Model Name and click the Multiple_Actions_button.pngmenu at the top of the list:
  3. Select Archive.

The models move from the Current tab to the Archive tab.

Important

To reinstate archived models, go to the Archive tab and select Actions > Reinstate.

When to regenerate models

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

Model_Evaluation.png

Evaluation is indicated as FairGood, or Excellent

Click on the Health_Details_button.png button to see details of the most recent evaluations. This can show you if the evaluation indicates a trend.

Model_Health.png

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

 

Tip: For faster searching, add an asterisk to the end of your partial query. Example: cert*