Editar

Compartir a través de

Optimize data migration for finance and operations apps

  • Artículo

Data migration is a key success factor in almost every implementation. A primary concern of some customers is the speed that data can be migrated at, especially if there are vast amounts of data and a small cutover window. The Data migration framework is also used to move data as part of business requirements and operations.

The information in this article represents a set of steps and actions that you can use to optimize the performance of data migration.

Note

Testing results in a Tier-1 environment should not be compared or extrapolated to performance in a Tier-2 or higher sandbox environment.

Not all standard entities have been optimized for data migration. Some entities have been optimized for integration with Open Data Protocol (OData), and if a required entity can't be optimized to meet the performance requirements, we recommend that you create a new optimized entity. A developer can accelerate this process by duplicating an existing entity.

Begin the optimization phase by using a subset of the data. For example, if you must import one million records, consider starting with 1,000 records, then increase the number to 10,000 records, and then increase it to 100,000 records.

After you've identified the entities that to use, you should go through the following sections to explore opportunities for optimization.

Disable change tracking

You can enable and disable change tracking from the list of entities.

  1. In the Data management workspace, select the Data entities tile.
  2. On the Target entities page, select the entity in the grid, and then, on the Action Pane, on the Change tracking tab, select Disable Change Tracking.

Enable set-based processing

Follow these steps to verify that an entity supports set-based processing.

  1. In the Data management workspace, select the Data entities tile.
  2. On the Target entities page, find the entity in the grid, and review the value in the Set-based processing column.

For an example that shows how set-based processing can be enabled for the General Journal entity, see Best practices for importing vouchers by using the General journal entity. Not all entities support set-based processing. For example, if you try to enable support set-based processing for the Customer definitions (CustCustomerBaseEntity) entity and save your change, you receive the following warning message:

Set operations not supported for 'Customer definitions' entity.

Here are some important considerations if you must create an entity that allows for set-based processing:

  • The data sources can't be read-only.
  • The ValidTimeStateEnabled property of the data entity view must be set to No.
  • All tables on the data sources must have TableType property set to Regular.
  • The property QueryType on the used Query can't be set to Union.
  • The main data source can't prevent data from being saved across companies. However, embedded data sources allow it.
  • The main data source can't prevent data from being saved across partitions. However, embedded data sources allow it.

Create a data migration batch group

During cutover, run the data migration while there's little or no other activity. It can be helpful to configure and use a batch group where most or all compute nodes are assigned.

You can configure a batch group on the Batch group page (System administration > Setup > Batch group).

Enable priority-based batch scheduling

In Platform update 31, the new priority-based batch scheduling feature optimizes how batch jobs are run. If contention is identified in the batch platform, consider enabling priority-based batch scheduling.

Configure the maximum number of batch threads

To better use parallelism and multithreading, you can configure the maximum number of batch threads per instance of Application Object Server (AOS) by setting the Maximum batch threads field on the Server configuration page (System administration > Setup > Server configuration). Be careful about changing the value of this field. A value that is too high can have negative performance implications. Currently, the default value is 8. You can increase the value to 12 or 16. However, you shouldn't set the field to a value that is more than 16 unless you do significant performance testing.

Import in batch mode

Whenever you run an import job, make sure that it's run in batch mode. Otherwise, a single thread is used to run the job. In this case, the system isn't able to take full advantage of these optimization configurations.

Clean staging tables

We recommend that you clean up the staging tables. You can achieve this optimization by scheduling the Job history cleanup job. To schedule this job, select the Job history cleanup tile in the Data management workspace.

Note

You must first turn on the Execution history cleanup feature in the Feature management workspace.

Update statistics

Before you run a data migration job that involves a large volume of data, consider updating the statistics across the associated tables. This optimization applies specifically to sandbox environments, because it's handled automatically in production environments. You can update statistics for a specific table from LCS. Alternatively, in a sandbox environment, you can use the sp_updatestats stored procedure through direct SQL.

Clean the data

The time that is spent on validations and error reporting increases the total time of the migration. Consider this fact when you import a high volume of invalid or inconsistent data. We recommend that you try to fix and reduce errors that are related to data quality. In this way, you help prevent unnecessary executions of validation and error handling.

Configurations to test during test runs of data migration

The following configurations can affect performance. Therefore, we recommend that you test changes by using different values that are suitable for your scenario.

Configure entity execution parameters

Follow these steps to change the execution parameters for all entities or specific entities.

  1. In the Data management workspace, select the Framework parameters tile.
  2. On the Data import/export framework parameters page, on the Entity settings tab, select Configure entity execution parameters.
  3. On the Entity import execution parameters page, set the Import threshold record count and Import task count fields as appropriate for the desired entities.

Import threshold record count

This value determines the number of records to be processed per thread. By modifying the Import threshold record count you can control how you want to split the import into smaller tasks.

Import task count

This field defines the number of threads that's used for the data migration job for a specific entity. For example, the Maximum batch threads field for each server is set to 8, and four servers are assigned to the data migration batch group. In this case, the total maximum value of the Import task count field is 32 (= 8 × 4).

If a data entity doesn't support multithreading, you receive an error message when you try to configure the entity. For example:

Custom sequence is defined for the entity 'Customers V3', more than one task is not supported.

Validations

Validation logic for record insertions or updates might have been incorporated into the system, or there might be validation of individual fields. If the data migration is mature enough, the time that is spent on imports can be reduced by disabling this validation, if it can be disabled.

Follow these steps to change the settings for each entity.

  1. In the Data management workspace, select the Data entities tile.
  2. On the Target entities page, select the entity in the grid, and then, on the Action Pane, select Entity structure.
  3. On the Entity structure page, set the Run business validations and Run business logic in insert or update method fields as appropriate.

Run business validations

If this check box is selected, the system runs any logic that is written into the validateWrite() method on the table. It also runs any related event handlers.

Run business logic in insert or update method

If this check box is selected, the system runs any logic that is written into the insert() or update() method on the table. It also runs any related event handlers.

Call the validateField method on the target

Follow these steps to run field validation.

  1. In the Data management workspace, select the Data entities tile.
  2. On the Target entities page, select the entity in the grid, and then, on the Action Pane, select Modify target mapping.
  3. On the Map staging to target page, select the Call validate Field method check box for the field that validation should be run for. The validateField(FieldId p1) method is then called for that field.

Recommendations for optimizing data migration performance

Here are some general recommendations about the approach that you should use to optimize the performance of data migration:

  • Break up large files into smaller chunks. This approach gives the SQL optimizer time to determine whether a new query plan is optimal.
  • Test performance in an appropriate Tier-2 or higher environment.
  • Test performance in a mock cutover before go-live.

Testing of data migration performance is an iterative process. We recommend that you collect and compare information about each test to determine the optimal configuration for a specific entity. You should collect and verify some of the following settings:

  • The batch group that is used
  • The number of batch servers that are assigned to each batch group
  • The maximum number of batch threads per batch server

Here is an example of the information that you might collect for each entity.

Information Description
Entity The name of the entity that is being tested
Number of records The number of records that are being imported
Source format The source format of the data that is being imported
Change tracking disabled Yes/No
Set-based processing Yes/No
Import threshold record count The number of records
Import task count The number of tasks
Run business validations Yes/No
Run business logic in insert or update method Yes/No
Call validate Field method Yes/No (potential field list)
Required performance The amount of time that the import must be completed within to achieve the cutover window
Actual performance The actual amount of time that was required to import records

There are more areas where performance can be optimized. For example, you can analyze the specific queries and query plans. However, those processes are covered in other topics.

For more information about performance troubleshooting and optimization, see the following topics: