FEFreeExamDumps.in

DP-800 Practice Questions — Page 6

Question 51

Open question ↗

Changes in an order table must trigger downstream processing in other Azure services. The integration should avoid tight coupling between the database and each consumer.

Which design should you recommend?

  • A.An event-driven pattern that publishes changes for downstream subscribers.
  • B.Synchronous calls from the database to every consumer service.
  • C.Manual email notifications from DBAs.
  • D.A nightly CSV export only.

Question 52

Open question ↗

You plan to expose selected database entities and stored procedures to web clients by using Data API Builder. Security requires least privilege and predictable contracts for both REST and GraphQL consumers.

Which action should you recommend?

  • A.Expose the entire database schema and rely on clients to use only approved tables.
  • B.Publish only the required entities and procedures, and map them to explicitly defined API operations.
  • C.Grant db_owner to the API identity.
  • D.Use server-level logins in client applications.

Question 53

Open question ↗

A development team is choosing between several model options for summarization, classification, and semantic retrieval over SQL data. They need a selection process tied to workload requirements rather than vendor preference.

Which factor is most important?

  • A.Match model capabilities to the actual SQL workload and evaluation criteria.
  • B.Choose the largest available model for every task.
  • C.Use the same model for all tasks without evaluation.
  • D.Prioritize model color theme consistency in dashboards.

Question 54

Open question ↗

A model used for SQL summarization has become more expensive, and the team is re-evaluating trade-offs between quality and cost.

Which process should you recommend?

  • A.Evaluate candidate models against the workload's quality and cost criteria.
  • B.Keep the current model forever without measurement.
  • C.Choose the newest model automatically.
  • D.Base the choice only on model name length.

Question 55

Open question ↗

A support portal uses embeddings generated from product manuals stored in Azure SQL Database. Updated manual sections must become searchable quickly without fully reprocessing all historical content.

Which strategy should you recommend?

  • A.Recompute embeddings only for rows that changed and track refresh metadata.
  • B.Drop the embeddings table on every update.
  • C.Store embeddings in a text file outside the database and reload them weekly.
  • D.Avoid keeping any metadata about embedding freshness.

Question 56

Open question ↗

A vector-enabled solution stores long support articles in SQL. The chunking strategy must preserve enough context for retrieval while avoiding oversized embeddings inputs.

Which action should you recommend?

  • A.Chunk content into semantically meaningful sections and keep metadata for reconstruction.
  • B.Embed the entire knowledge base as one row.
  • C.Discard all section metadata.
  • D.Use random chunk boundaries only.

Question 57

Open question ↗

A vector-enabled solution stores long support articles in SQL. The chunking strategy must preserve enough context for retrieval while avoiding oversized embeddings inputs.

Which action should you recommend?

  • A.Chunk content into semantically meaningful sections and keep metadata for reconstruction.
  • B.Embed the entire knowledge base as one row.
  • C.Discard all section metadata.
  • D.Use random chunk boundaries only.

Question 58

Open question ↗

An embeddings table must support incremental refresh after source rows change.

Which data point should you track?

  • A.A last-embedded timestamp or version marker per row
  • B.The user's monitor resolution
  • C.A random note field
  • D.Printer status

Question 59

Open question ↗

A customer portal supports both exact product-code lookups and natural-language troubleshooting queries. The search solution must support keyword precision and semantic recall.

Which approach should you recommend?

  • A.Use only full-text search.
  • B.Use only vector similarity without metadata filters.
  • C.Use a hybrid search strategy that combines lexical and vector signals.
  • D.Store search results in screenshots.

Question 60

Open question ↗

A hybrid search solution must combine product-code precision with semantic troubleshooting help.

Which ranking design should you recommend?

  • A.Combine lexical and vector signals in the ranking process.
  • B.Use semantic similarity only for all searches.
  • C.Remove filters before ranking.
  • D.Sort only by document creation date.