A reporting team needs to join lakehouse order tables against a customer table that lives in an operational PostgreSQL database. The customer table changes constantly and the join runs a few times a day. What is the appropriate approach in watsonx.data?

The appropriate approach is to federate the PostgreSQL source so the join reads it in place, keeping the customer data current without a copy. Federation lets an engine query a remote source in place through a connector, so the join sees the current state of the customer table without any movement. With a few executions a day and a small dimension table, the cost of reaching across the network is modest and the freshness benefit is real.

Scheduling a nightly full copy of the customer table into the lakehouse and joining against the copy is incorrect because it guarantees the reports are up to twenty-four hours behind on data the team specifically described as constantly changing. It also introduces a pipeline, a schedule and a failure mode for no gain.

Exporting the customer table to CSV before each run and uploading it to the bucket is incorrect because manual export and upload is the same staleness problem with added human effort, and the resulting files sit outside the governance that applies to the source.

Replicating the entire PostgreSQL instance into object storage so all its tables are available is incorrect because it is a large amount of movement, storage and ongoing synchronisation for a requirement that federation covers directly.

For the exam, the trigger words for federation are cross-source, current, and avoid unnecessary movement.