I have a system landscape consisting of multiple applications.

• Each application provides REST interfaces to communicate with the outside world.
• Each application has its own data storage (some kind of database).
• Each application writes messages for other applications to consume.
• Each application builds its read model of the others application's data by consuming the message streams.

I now need to populate the system landscape with test data. There would be two ways to do this:

• Create the test data via the REST interfaces (like it would be done in production too)
  One disadvantage would be that for the test data to be consistent, I would need to wait for the response of each application so that I can then create dependent test data in the next application.
  Another disadvantage would be that the other applications need to be able to handle a flood of messages that would be created when the test data is created.
• Directly load the test data via SQL or another route into the read model of each application (bypassing the REST interface)
  This would be faster and would not create a flood of messages.

Are there any major drawbacks to the second approach? Are there best practices for loading test data into a distributed application landscape?

I have a system landscape consisting of multiple applications.

• Each application provides REST interfaces to communicate with the outside world.
• Each application has its own data storage (some kind of database).
• Each application writes persistent messages for other applications to consume.
• Each application builds its read model of the others application's data by consuming these message streams.

I now need to populate the system landscape with test data. There would be two ways to do this:

• Create the test data via the REST interfaces (like it would be done in production too)
  One disadvantage would be that for the test data to be consistent, I would need to wait for the response of each application so that I can then create dependent test data in the next application.
  Another disadvantage would be that the other applications need to be able to handle a flood of messages that would be created when the test data is created.
• Directly load the test data via SQL or another route into the read model of each application (bypassing the REST interface)
  This would be faster and would not create a flood of messages.

Are there any major drawbacks to the second approach? Are there best practices for loading test data into a distributed application landscape?