You are using a Dataflow streaming job to read messages from a message bus that does not support exactly-once delivery. Your job then applies some transformations, and loads the result into…

Question

Options

• AUse the BigQuery Storage Write API and ensure that your target BigQuery table is regional.
• BUse the BigQuery Storage Write API and ensure that your target BigQuery table is multiregional.
• CUse the BigQuery Streaming API and ensure that your target BigQuery table is regional.
• DUse the BigQuery Streaming API and ensure that your target BigQuery table is multiregional.

• A
  76% (44)
• B
  7% (4)
• C
  3% (2)
• D
  14% (8)

Explanation

Option A is correct because the BigQuery Storage Write API is the only option that provides true exactly-once delivery semantics - it uses stream offsets and committed write streams to prevent duplicate records, unlike the legacy Streaming API. Regional BigQuery tables are required here because the Storage Write API has a default throughput quota of 3 GB/s for regional datasets but only 1 GB/s for multiregional datasets (US/EU), making regional the only choice that handles the expected 1.5 GB/s load.

Option B fails because the multiregional quota cap of ~1 GB/s would be exceeded at 1.5 GB/s throughput. Options C and D both fail because the legacy BigQuery Streaming API provides at-least-once delivery with best-effort deduplication - not the guaranteed exactly-once semantics the question requires.

Memory tip: Think "Storage Write = Strictly-once" - the "S" links Storage Write to exactly-once guarantees. Then remember that higher throughput demands require regional (not multiregional) placement, just as a regional data center is more focused/specialized than a spread-out multiregional one.

No community discussion yet for this question.