Microsoft DP-700 Practice Test 20

Question 190

IoT records must be queryable within seconds, but current nightly Copy jobs deliver them the next day. What loading architecture better matches the latency target?

Move the once-nightly Copy trigger to early morning and retain the same interval.
Enable native Delta file compaction but keep the nightly source acquisition schedule.
Implement a supported continuous ingestion route, such as Eventstream to an Eventhouse or compatible streaming destination, sized to the source and query needs.
Use a KQL visualization over yesterday's warehouse import and call it near-real-time.

Explanation

Correct Answer

A live event path reduces acquisition latency compared with batch scheduling and can expose recently received events to KQL queries.

Incorrect Answers

Changing the batch hour doesn't satisfy a seconds-level event-to-query requirement.

Faster scans won't make source events available before the next batch arrives.

A real-time query interface can't provide new messages that are still waiting for nightly ingestion.

Question 191

A team promises sub-minute telemetry visibility but still relies on a daily Copy job. Why does it fail the requirement?

The Eventhouse table is poorly optimized for query acceleration.
The telemetry records lack a sensitivity label and therefore arrive one day late.
The IoT source needs a larger warehouse row group to meet query freshness.
The batch trigger interval is much longer than the latency objective; a continuous event ingestion mechanism is missing.

Explanation

Correct Answer

The chosen architecture inherently waits for the next scheduled batch, regardless of how quickly the Copy activity executes.

Incorrect Answers

Indexing or caching might slow queries after ingest, but a daily Copy schedule fundamentally prevents new events from arriving at sub-minute cadence.

Classification metadata cannot change the delay imposed by a job that runs only once per day.

Warehouse storage layout is independent of whether telemetry is delivered continuously. A streaming ingestion path is needed.

Question 192

After switching to continuous ingestion, what evidence validates the latency objective?

Emit timestamped test events and measure source emission to successful destination-query visibility under representative load.
Confirm the Eventstream-to-Eventhouse topology exists but send no sample telemetry.
Measure Eventhouse query execution time only, excluding the source-to-ingest delay.
Use nightly row counts to claim a seconds-level event arrival target.

Explanation

Correct Answer

Arrival-to-query time measures the actual service level and can expose processing queues that a successful connector handshake does not.

Incorrect Answers

A configured route can be paused, broken or delayed. Validate real timestamped records reaching the query surface.

Fast KQL execution does not prove fresh data. Measure source emission to destination query visibility against the SLA.

Daily counts cannot resolve per-event propagation latency, especially at production-like throughput.

Question 193

A Fabric ingestion job can be retried after network interruptions. The destination must never duplicate a customer transaction key. Which sink pattern is most suitable?

Generate a fresh destination transaction ID every time the same source event is retried.
Use an idempotent merge/upsert keyed by stable transaction ID, with a compatible transaction or commit boundary for retries.
Increase activity retry attempts without making target writes idempotent.
Remove the transaction business key before the merge to improve throughput.

Explanation

Correct Answer

Repeating the same input batch then updates or matches existing keys instead of blindly appending duplicate rows.

Incorrect Answers

Changing identifiers prevents duplicate recognition and can multiply business transactions after interruptions.

More retries can amplify duplication when the sink blindly appends each attempt.

Stable keys are necessary to match previously applied records during retry-safe upserts.

Question 194

A Copy job is retried with the same partition and target count doubles. The sink is append-only without key reconciliation. What caused it?

The source emitted duplicated data because its business transaction identifiers were stable.
Copy automatically converts all repeat deliveries into updates in Append write mode.
The retry re-appended records that were already committed because the load is not idempotent on stable business identifiers.
A dashboard displayed the same rows twice because KQL query results were cached.

Explanation

Correct Answer

At-least-once execution needs duplicate-safe sink logic; a second successful append is not automatically recognized as replay.

Incorrect Answers

Stable keys make deduplication possible; the defect is blind append on retry, which does not match previously committed keys.

Append creates new rows. Automatic upsert is not supplied unless explicitly configured and supported by the sink.

The target count actually doubled after retry. That is a write-path idempotency issue, not only query presentation.

Question 195

An ETL team introduces key-based merging to prevent retry duplication. What test should it run?

Run two batches with different transaction IDs and confirm the target contains both.
Check only the Merge activity returns Success for each invocation.
Compare the target total count once without checking matching keys or updates.
Execute an identical source batch twice and confirm the target contains one correct row per transaction key with expected updates and unchanged totals.

Explanation

Correct Answer

Rerun testing directly validates the durable idempotence property rather than relying on the absence of routine failures.

Incorrect Answers

Distinct keys should produce distinct records; this does not test duplicate-safe behavior for an identical retried batch.

Both runs may succeed even when the merge predicate is wrong and creates duplicate business keys. The target invariant must be checked.

One count can conceal duplicate keys and missed corrections. Exact per-key cardinality and expected value behavior are required.

Question 196

A bronze-to-silver data pipeline must support replay when business validation rules change. Which storage design preserves that ability?

Retain a trustworthy raw bronze landing record and produce separately validated silver output with reproducible transformations.
Overwrite the bronze raw-event store with each newly approved silver transformation.
Retain only aggregate counts from ingestion and discard unvalidated source columns.
Store only the filtered silver output and infer any rejected rows from dashboard totals.

Explanation

Correct Answer

A raw replay boundary preserves the original ingested evidence so transformations can be corrected without needing to recollect the source.

Incorrect Answers

Destroying the original source prevents reprocessing when the validation rules change.

Counts cannot reconstruct the original records needed for new cleansing or business rules.

Without immutable or replayable bronze inputs, rejected records and newly relevant fields cannot be recovered.

Question 197

A team overwrites its only raw input files after each silver transformation. Later it cannot reproduce old ingestion results. Why?

The silver output is versioned so the raw source can always be reconstructed exactly.
It destroyed the source-of-truth replay inputs, leaving only previously transformed data with no original baseline.
The pipeline’s sensitivity label was lost during a transformation run.
The source changed because the team ran a read-only SQL query against a warehouse view.

Explanation

Correct Answer

Without preserved raw content and lineage, historical reprocessing cannot reconstruct records that were filtered or modified in the original transformation.

Incorrect Answers

A transformed table can be lossy; source records and original fields may have been filtered, normalized or discarded. Silver does not replace retained bronze inputs.

Classification metadata is unrelated to deterministic replay when the original input bytes were overwritten.

Reading a downstream view does not erase the raw ingestion baseline. The overwrite of the only original input is the causal event.

Question 198

A bronze/silver architecture claims to support deterministic replay. What experiment proves it?

Run the current live source again and assume it is byte-identical to the earlier bronze batch.
Compare only that a new silver table was created after a rerun.
Reprocess a frozen bronze batch with a pinned transformation version into an isolated silver target and compare keys, values and validation outcomes.
Change the pipeline run name and check whether the stored silver rows remain visible.

Explanation

Correct Answer

A controlled replay can establish reproducibility and identify nondeterministic transformations rather than assuming the architecture is recoverable.

Incorrect Answers

Live source data can change. Replay verification requires a frozen input snapshot and a pinned transform version.

Table creation is not reproducibility evidence. Relevant keys, values and validation results must match the expected replay contract.

A different name does not prove the raw batch was preserved or the transformation gives deterministic results.

Question 199

A historical order amount was corrected in the source after the day was already loaded into a Fabric warehouse. What loading technique updates the existing fact?

Append the corrected order as a second current fact row with a new surrogate key.
Ignore source corrections whose OrderDate precedes yesterday.
Rebuild every dimension table rather than matching the corrected fact's existing business key.
Use correction-aware key-based MERGE or an equivalent supported update path to reconcile historical facts, not only append newly dated records.

Explanation

Correct Answer

Stable business identifiers permit existing target facts to be revised while avoiding a duplicate record for the same logical transaction.

Incorrect Answers

That keeps both amounts for one business transaction and inflates reporting unless the history model explicitly requires versions.

The corrected historical order is intentionally older; date-only exclusion would leave the fact wrong.

Dimension rebuilds are unrelated to targeted correction of an already-loaded fact amount.

Question 200

An ETL job appends a revised historical order under the same order key, leaving both old and new amounts. What defect caused it?

The target load treats the correction as an unrelated insert instead of matching and updating the existing business key.
The source correction was assigned the same stable order key, so MERGE cannot distinguish it.
The historical order was intentionally preserved by Type 2 versioning with valid effective dates.
V-Order caused a duplicate SQL fact to appear after physical compaction.

Explanation

Correct Answer

A correction requires key-based reconciliation or intentional versioning; append-only handling duplicates a fact when historical rows are mutable.

Incorrect Answers

A stable key is exactly what enables a correction-aware update. The pipeline instead treats the incoming row as an unrelated append.

Version history could be a legitimate policy, but the scenario describes uncoordinated duplicate amounts under the same order key without the specified Type 2 contract.

V-Order reorganizes physical storage; it does not create new logical rows with the same order key. The append write path did.

Implementing Data Engineering Solutions Using Microsoft Fabric (DP-700) Practice Tests

Microsoft DP-700 Practice Test 18 - 10 Questions (Last Reviewed 17-Oct-2026)
Microsoft DP-700 Practice Test 19 - 10 Questions (Last Reviewed 16-Oct-2026)
Microsoft DP-700 Practice Test 20 - 10 Questions (Last Reviewed 15-Oct-2026)
Microsoft DP-700 Practice Test 21 - 10 Questions (Last Reviewed 14-Oct-2026)
Microsoft DP-700 Practice Test 22 - 10 Questions (Last Reviewed 13-Oct-2026)

Pass Your Microsoft DP-700 Exam Easy!

Question 190

Explanation

Correct Answer

Incorrect Answers

Question 191

Explanation

Correct Answer

Incorrect Answers

Question 192

Explanation

Correct Answer

Incorrect Answers

Question 193

Explanation

Correct Answer

Incorrect Answers

Question 194

Explanation

Correct Answer

Incorrect Answers

Question 195

Explanation

Correct Answer

Incorrect Answers

Question 196

Explanation

Correct Answer

Incorrect Answers

Question 197

Explanation

Correct Answer

Incorrect Answers

Question 198

Explanation

Correct Answer

Incorrect Answers

Question 199

Explanation

Correct Answer

Incorrect Answers

Question 200

Explanation

Correct Answer

Incorrect Answers

Implementing Data Engineering Solutions Using Microsoft Fabric (DP-700) Practice Tests