Oracle Data Loader On Demand Guide > Overview of Oracle Data Loader On Demand >

Optimizing Load Size

When performing a large load, keep in mind that the entire environment might need to be sized appropriately for the load that you plan to perform. There are many factors to take into consideration to ensure optimal results.

To ensure optimal load size

1. Check the client configuration to ensure data is sent at an optimal speed.
2. Inspect the data to avoid errors.

   You might want to load your data in a particular sequence because one record might depend on another record's existence. For example, if you want to import your users, and your data has a Reports To field, you might want to import users in hierarchical order so the executives are already in the database before their subordinates. These records cannot be imported randomly, but require sequential processing.

3. Check for other data quality issues, such as the required fields and pick list values are valid.
4. Make sure that the log level is tailored to the specific load.

   There is no need for unnecessary logging.

   Some recurring data loads might need to be suspended by Oracle CRM On Demand Hosting Operations personnel, and they might need to restrict the use of the environment.

5. Perform test runs and collect metrics for each test run. Scale or reconfigure the environment iteratively to achieve the required throughput rate.

   Optimizing typically requires a few weeks because it is an iterative process of plan, test, analyze, adjust, and repeat.

Factors That Affect Performance When Performing a Large Load

The following information relates to performance and sizing concerns:

• Throughput characteristics between different record types and scenarios vary widely, therefore it is not possible to predict the throughput rate for a given load.
• Optimization is an iterative process which builds from a single session test to large scale concurrency testing.
• It is your responsibility to act on your data. Oracle does not update, delete, or otherwise alter customer data.
• Staging and production environments are not identical, therefore throughput can differ between these two environments.
• Oracle Data Loader On Demand uses nonsequential processing. This provides for maximum throughput by using all available resources to import your data. For example, if your server has four tasks available and you submitted a 4000-record import, and if no one else was importing data, then Oracle Data Loader On Demand would use all four tasks to import the 4000 records (each task importing 1000 records). If this job was processed using a sequential process, then only one task would be used and the other three would be idle.
• For better performance when importing, consider inactivating any workflows that can be postponed.