McAfee Secure

Exam Code: DP-800

Exam Name: Developing AI-Enabled Database Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: SQL AI Developer Associate

Microsoft DP-800 Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

101 Questions & Answers with Testing Engine

"Developing AI-Enabled Database Solutions Exam", also known as DP-800 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date DP-800 Exam Engine. Your DP-800 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft DP-800 Practice Test 8

Question 70

A reporting transaction under READ COMMITTED sees changing totals between consecutive statements. Which measurable outcome BEST verifies that the requirement is satisfied?

  1. A stale client receives a conflict and cannot overwrite the newer row.
  2. Repeated statements in one snapshot transaction give a consistent view.
  3. A before/after sample reduces CPU without harming critical analytics.
  4. A forced or corrected plan improves tested workload without wrong results.

Explanation

Correct Answer

B:

Repeated statements in one snapshot transaction give a consistent view. This is the decisive observable evidence after implementing evaluate SNAPSHOT isolation or appropriate transaction design for repeatable point-in-time reads. It tests the underlying requirement: Isolation choices define which committed versions are visible across statements.

Incorrect Answers

A:

A stale client receives a conflict and cannot overwrite the newer row. However, that result would not establish the required outcome: repeated statements in one snapshot transaction give a consistent view.

C:

A before/after sample reduces CPU without harming critical analytics. However, that result would not establish the required outcome: repeated statements in one snapshot transaction give a consistent view.

D:

A forced or corrected plan improves tested workload without wrong results. However, that result would not establish the required outcome: repeated statements in one snapshot transaction give a consistent view.

Question 71

An analyst wants to use ordinary LIKE predicates against randomized Always Encrypted ciphertext. Which technical reason BEST explains the proposed correction?

  1. Column key hierarchy and client metadata must agree for encrypted column access.
  2. Agents can turn natural-language requests into real database mutations if tools permit.
  3. Randomized encryption does not support general plaintext pattern matching inside the engine.
  4. Untrusted headers cannot be treated as authenticated authorization claims.

Explanation

Correct Answer

C:

Randomized encryption does not support general plaintext pattern matching inside the engine. This explains why the corrective action is to redesign the query or use an explicitly supported encryption/enclave pattern with documented tradeoffs. Confirm the diagnosis when supported equality or enclave test cases behave as designed without broad plaintext exposure.

Incorrect Answers

A:

Column key hierarchy and client metadata must agree for encrypted column access. The actual technical constraint is that randomized encryption does not support general plaintext pattern matching inside the engine.

B:

Agents can turn natural-language requests into real database mutations if tools permit. The actual technical constraint is that randomized encryption does not support general plaintext pattern matching inside the engine.

D:

Untrusted headers cannot be treated as authenticated authorization claims. The actual technical constraint is that randomized encryption does not support general plaintext pattern matching inside the engine.

Question 72

A SQL service's identity can call a model but lacks permission to read a required secure search index. What maintenance or governance control BEST reduces recurrence of this problem?

  1. Scope endpoint permissions and monitor request consumption.
  2. Reset pooled connections and validate context on each request.
  3. Separate key custodianship and audit key usage.
  4. Review grants separately when components are replaced.

Explanation

Correct Answer

D:

Review grants separately when components are replaced. This guards against recurrence because identity permission at one Azure service does not imply access to adjacent resources. The direct remediation is: Configure distinct least-privilege grants on model and search resources.

Incorrect Answers

A:

Scope endpoint permissions and monitor request consumption. The current task instead requires this action: review grants separately when components are replaced.

B:

Reset pooled connections and validate context on each request. The current task instead requires this action: review grants separately when components are replaced.

C:

Separate key custodianship and audit key usage. The current task instead requires this action: review grants separately when components are replaced.

Question 73

A recursive organization-chart query runs indefinitely after cyclic parent links are introduced. Which implementation decision BEST addresses this requirement?

  1. Validate acyclic parent relationships and bound recursion depth.
  2. Use a recursive CTE with an anchor member and a terminating recursive member.
  3. Include an explicit window ORDER BY with unique tie-breaker columns.
  4. Use a correlated aggregate subquery or window function scoped by category.

Explanation

Correct Answer

A:

Validate acyclic parent relationships and bound recursion depth. This targets the primary issue because a cycle can repeatedly regenerate the same nodes in a recursive CTE. Verification: A cyclic test input is detected rather than consuming unbounded resources. Prevent recurrence: Implement graph integrity checks on writes.

Incorrect Answers

B:

Using a recursive CTE with an anchor member and a terminating recursive member would not correct a recursive organization-chart query runs indefinitely after cyclic parent links are introduced. Validate acyclic parent relationships and bound recursion depth.

C:

Window ranking order is defined by its OVER clause, not the outer result sort.

D:

An outer reference lets the inner aggregate vary by each row's category.

Question 74

A dev wants to enable all query optimization flags globally because one report is slow. Which measurable outcome BEST verifies that the requirement is satisfied?

  1. A peak replay meets throughput and latency targets within budget.
  2. The report improves without regression in representative unrelated queries.
  3. A stale client receives a conflict and cannot overwrite the newer row.
  4. Correctness tests pass without excessive lock waits under the selected isolation.

Explanation

Correct Answer

B:

The report improves without regression in representative unrelated queries. This is the decisive observable evidence after implementing use targeted measured configuration changes and a rollback plan rather than blindly enabling trace flags. It tests the underlying requirement: Database-level settings can affect many workloads and engine versions differently.

Incorrect Answers

A:

A peak replay meets throughput and latency targets within budget. However, that result would not establish the required outcome: the report improves without regression in representative unrelated queries.

C:

A stale client receives a conflict and cannot overwrite the newer row. However, that result would not establish the required outcome: the report improves without regression in representative unrelated queries.

D:

Correctness tests pass without excessive lock waits under the selected isolation. However, that result would not establish the required outcome: the report improves without regression in representative unrelated queries.

Question 75

A query uses vectors with unequal magnitudes and wants unit-length normalized comparisons. Which technical reason BEST explains the proposed correction?

  1. Vector metadata functions describe representation without parsing serialized JSON arrays.
  2. Native vector types validate dimensions and avoid repeated parsing of text arrays.
  3. Normalizing vector length separates direction from magnitude for suitable similarity metrics.
  4. Lexical retrieval protects exact tokens while vectors address paraphrases.

Explanation

Correct Answer

C:

Normalizing vector length separates direction from magnitude for suitable similarity metrics. This explains why the corrective action is to use VECTOR_NORMALIZE where supported before the intended comparison workflow. Confirm the diagnosis when normalized test vectors have unit norm within floating tolerance.

Incorrect Answers

A:

Vector metadata functions describe representation without parsing serialized JSON arrays. The actual technical constraint is that normalizing vector length separates direction from magnitude for suitable similarity metrics.

B:

Native vector types validate dimensions and avoid repeated parsing of text arrays. The actual technical constraint is that normalizing vector length separates direction from magnitude for suitable similarity metrics.

D:

Lexical retrieval protects exact tokens while vectors address paraphrases. The actual technical constraint is that normalizing vector length separates direction from magnitude for suitable similarity metrics.

Question 76

A salesperson needs to compare each month's sales with the prior month's result. What maintenance or governance control BEST reduces recurrence of this problem?

  1. Review partitions and tie order when requirements change.
  2. Guard against cyclic paths and unsupported graph features.
  3. Enforce ownership and governance over graph writes.
  4. Specify missing predecessor behavior explicitly.

Explanation

Correct Answer

D:

Specify missing predecessor behavior explicitly. This guards against recurrence because LAG references a preceding ordered row without self-joining every record. The direct remediation is: Use LAG(SalesAmount) OVER(PARTITION BY Region ORDER BY SalesMonth).

Incorrect Answers

A:

Review partitions and tie order when requirements change. The current task instead requires this action: specify missing predecessor behavior explicitly.

B:

Guard against cyclic paths and unsupported graph features. The current task instead requires this action: specify missing predecessor behavior explicitly.

C:

Enforce ownership and governance over graph writes. The current task instead requires this action: specify missing predecessor behavior explicitly.

Question 77

A service trusts an unsigned client-supplied user role in a REST header. What should the SQL developer do FIRST to address the specific issue?

  1. Validate provider-issued identity tokens and map claims to authorized application roles.
  2. Configure authentication and per-entity role/operation permissions rather than public wildcard actions.
  3. Remove keys and use deployment-time secret references or supported managed identity.
  4. Require approved identity authentication and network controls for the inference endpoint.

Explanation

Correct Answer

A:

Validate provider-issued identity tokens and map claims to authorized application roles. This targets the primary issue because untrusted headers cannot be treated as authenticated authorization claims. Verification: A forged admin header fails while valid signed identities receive correct rights. Prevent recurrence: Rotate signing keys and validate audiences and issuers.

Incorrect Answers

B:

API generation does not automatically imply that exposed data is appropriately authorized.

C:

Repository history retains secrets even after a visible commit is removed.

D:

An open endpoint may expose inputs, outputs or costly compute to unauthorized callers.

Question 78

A developer tries to use CONTAINS on a text column without a full-text index. Which acceptance check is MOST directly related to this scenario?

  1. The quoted phrase test excludes documents without that phrase.
  2. A controlled CONTAINS query succeeds after population completes.
  3. Recall@10 and p95 latency meet separately defined thresholds.
  4. Known nearest-neighbor fixtures appear first in the final order.

Explanation

Correct Answer

B:

A controlled CONTAINS query succeeds after population completes. This is the decisive observable evidence after implementing create a supported full-text index and required unique key on the searchable table. It tests the underlying requirement: Full-text predicates require a configured index and supported column types.

Incorrect Answers

A:

The quoted phrase test excludes documents without that phrase. However, that result would not establish the required outcome: a controlled CONTAINS query succeeds after population completes.

C:

Recall@10 and p95 latency meet separately defined thresholds. However, that result would not establish the required outcome: a controlled CONTAINS query succeeds after population completes.

D:

Known nearest-neighbor fixtures appear first in the final order. However, that result would not establish the required outcome: a controlled CONTAINS query succeeds after population completes.

Question 79

An auditor sees database logs kept on the same instance a compromised admin can delete. Which technical property or constraint MOST directly matters here?

  1. Audit collection adds actor, operation and time evidence but does not itself block changes.
  2. Exposing a constrained view can limit data surface without broad base-table rights.
  3. Tamper-resistant off-host retention is stronger evidence than only local logs.
  4. Dynamic SQL may execute under a permission context different from static ownership chains.

Explanation

Correct Answer

C:

Tamper-resistant off-host retention is stronger evidence than only local logs. This explains why the corrective action is to send audit output to appropriately protected centralized storage with restricted permissions. Confirm the diagnosis when a simulated service administrator cannot erase the retained central record.

Incorrect Answers

A:

Auditing collection adds actor, operation and time evidence but does not itself block changes concerns a different mechanism; the relevant technical point is that tamper-resistant off-host retention is stronger evidence than only local logs.

B:

Exposing a constrained view can limit data surface without broad base-table rights. The actual technical constraint is that tamper-resistant off-host retention is stronger evidence than only local logs.

D:

Dynamic SQL may execute under a permission context different from static ownership chains. The actual technical constraint is that tamper-resistant off-host retention is stronger evidence than only local logs.

Question 80

A finance application needs deterministic machine-readable response fields from model inference. Which preventive practice BEST addresses the underlying risk over time?

  1. Track model quality drift and escalation thresholds.
  2. Keep source citations and access metadata attached.
  3. Evaluate per-language drift and fallback routing.
  4. Reject unexpected fields and never trust generated totals without validation.

Explanation

Correct Answer

D:

Reject unexpected fields and never trust generated totals without validation. This guards against recurrence because unconstrained prose is fragile to parse and can omit required keys. The direct remediation is: Prefer a model and API mode supporting validated structured output and enforce schema checks.

Incorrect Answers

A:

Track model quality drift and escalation thresholds. The current task instead requires this action: reject unexpected fields and never trust generated totals without validation.

B:

Keep source citations and access metadata attached. The current task instead requires this action: reject unexpected fields and never trust generated totals without validation.

C:

Evaluate per-language drift and fallback routing. The current task instead requires this action: reject unexpected fields and never trust generated totals without validation.

Developing AI-Enabled Database Solutions (DP-800) Practice Tests