Question

An e-commerce team needs to ingest 500,000 order documents nightly into a Cosmos DB container partitioned by /customerId. The documents span thousands of different customerId values, and each document is independent — partial failures are acceptable and should simply be retried. The team wants maximum ingestion throughput with the least code complexity. Which approach should they use?

Accepted Answer

Enable bulk execution (AllowBulkExecution = true) in the SDK and issue concurrent CreateItemAsync calls

Answer

Group the documents and use TransactionalBatch operations across the different customerId values

Answer

Wrap all inserts in a single stored procedure executed once per night

Answer

Insert documents sequentially with individual CreateItemAsync calls to preserve ordering