Microsoft DP-900 Practice Test 114
Question 1130
A factory needs alerts seconds after machine vibration crosses a threshold. Which choice BEST meets this requirement?
- Separate descriptive measurement from diagnostic analysis of contributing factors.
- Establish shared metric definitions and validate semantic calculations.
- Ingest and evaluate events as they arrive with a streaming pipeline.
- Run one report at midnight and label it a continuous stream.
Explanation
Correct Answer
C:Streaming processes ongoing events instead of waiting for a full batch.
Incorrect Answers
A:Separate descriptive measurement from diagnostic analysis of contributing factors. That action primarily handles Descriptive versus diagnostic analytics; the given requirement is Event streaming continual processing. Streaming processes ongoing events instead of waiting for a full batch.
B:Establish shared metric definitions and validate semantic calculations. That action primarily handles Data analyst metric definition; the given requirement is Event streaming continual processing. Streaming processes ongoing events instead of waiting for a full batch.
D:Run one report at midnight and label it a continuous stream. It does not implement Event streaming continual processing. Treating a high-rate stream as an hourly CSV export delays detection.
Question 1131
A vendor wants shared provisioned compute across multiple Azure SQL databases. During testing, the team observes this specific failure: A single noisy tenant can consume shared capacity without proper governance. Which corrective check should be performed FIRST?
- Combine all databases into one CSV and call it an elastic pool.
- Evaluate supported read scale-out options under the chosen service tier.
- Specify ORDER BY with a deterministic tie-breaker when paging.
- Monitor per-database and pool resource metrics under load.
Explanation
Correct Answer
D:Monitor per-database and pool resource metrics under load. A successful correction should demonstrate that peak tenants borrow capacity while pool-wide utilization stays manageable.
Incorrect Answers
A:Combine all databases into one CSV and call it an elastic pool. This reinforces the error rather than addressing the Azure SQL elastic pool sharing failure. Elastic pools support shared resource budgeting for multiple databases.
B:Evaluate supported read scale-out options under the chosen service tier. It serves Azure SQL read scale and reporting replicas, not the diagnosed Azure SQL elastic pool sharing problem. A single noisy tenant can consume shared capacity without proper governance.
C:Specify ORDER BY with a deterministic tie-breaker when paging. It serves SQL ORDER BY pagination stability, not the diagnosed Azure SQL elastic pool sharing problem. A single noisy tenant can consume shared capacity without proper governance.
Question 1132
A transaction fails after creating the order header but before allocating stock. Which principle BEST explains the correct OLTP atomic purchase transaction design decision?
- Transactional workloads emphasize consistent low-latency create update delete operations.
- CSV is a delimited flat-text format not a hierarchical JSON object.
- Flexible records may omit fields while preserving other consistent identifiers.
- Latency measures time per operation; throughput measures completed work rate.
Explanation
Correct Answer
A:Transactional workloads emphasize consistent low-latency create update delete operations. In this case the applicable decision is to keep related short business writes in an atomic transaction.
Incorrect Answers
B:CSV is a delimited flat-text format not a hierarchical JSON object. That concept explains CSV delimited text schema; the situation instead concerns OLTP atomic purchase transaction. Partial independent commits can oversell inventory.
C:Flexible records may omit fields while preserving other consistent identifiers. That concept explains Sparse document attributes; the situation instead concerns OLTP atomic purchase transaction. Partial independent commits can oversell inventory.
D:Latency measures time per operation; throughput measures completed work rate. That concept explains Operational SLA latency versus throughput; the situation instead concerns OLTP atomic purchase transaction. Partial independent commits can oversell inventory.
Question 1133
Batch transformations exceed capacity of a single local script. What is the MOST appropriate next validation or correction?
- Prototype one Spark pipeline and one reporting query.
- Rename a Python list spark and assume calculations distribute automatically.
- Test unique keys and count fact matches per dimension row.
- Verify Spark job execution and source-to-result reconciliation.
Explanation
Correct Answer
D:Verify Spark job execution and source-to-result reconciliation. This tests the intended result: distributed transformations produce correct partitions and aggregated outputs.
Incorrect Answers
A:Prototype one Spark pipeline and one reporting query. It tests Lakehouse versus Warehouse selection rather than the critical Azure Databricks Apache Spark processing behavior. Spark processes distributed datasets in parallel across compute workers.
B:Rename a Python list spark and assume calculations distribute automatically. This unsafe shortcut leaves the Azure Databricks Apache Spark processing problem unresolved. Expecting a notebook file alone to execute a cluster without compute is incorrect.
C:Test unique keys and count fact matches per dimension row. It tests One-to-many semantic relationships rather than the critical Azure Databricks Apache Spark processing behavior. Spark processes distributed datasets in parallel across compute workers.
Question 1134
Finance needs a curated history of sales measures by fiscal month. If the relevant Analytical warehouse for structured reporting safeguard is skipped, which failure is MOST directly associated with that omission?
- Treating production checkout row updates as a warehouse's primary workload can hurt analytics design.
- Assuming any flat blob container has identical POSIX-style directory operations can break workflows.
- Partial independent commits can oversell inventory.
- Skipping input validation contaminates downstream metrics.
Explanation
Correct Answer
A:Treating production checkout row updates as a warehouse's primary workload can hurt analytics design. A warehouse organizes integrated analytical tables optimized for reporting. A suitable check is to validate grain measures and aggregations with sample fiscal periods.
Incorrect Answers
B:Assuming any flat blob container has identical POSIX-style directory operations can break workflows. This failure is associated with Azure Data Lake Storage Gen2 analytics landing; it is not the primary failure in the stated Analytical warehouse for structured reporting scenario. A warehouse organizes integrated analytical tables optimized for reporting.
C:Partial independent commits can oversell inventory. This failure is associated with OLTP atomic purchase transaction; it is not the primary failure in the stated Analytical warehouse for structured reporting scenario. A warehouse organizes integrated analytical tables optimized for reporting.
D:Skipping input validation contaminates downstream metrics. This failure is associated with ETL data transformation before loading; it is not the primary failure in the stated Analytical warehouse for structured reporting scenario. A warehouse organizes integrated analytical tables optimized for reporting.
Question 1135
A fraud system must detect unusual card events within seconds. Which choice BEST meets this requirement?
- Configure an appropriate data gateway and scheduled refresh where needed.
- Apply a justified real-time anomaly rule with quality checks and alert thresholds.
- Call a once-daily CSV read a continuous real-time stream.
- Select streaming for continuous low-latency needs and batch for bounded periodic jobs.
Explanation
Correct Answer
D:Select streaming for continuous low-latency needs and batch for bounded periodic jobs. Streaming processes arriving records continually unlike scheduled batch. In the stated case, verification should show that fraud alerts arrive promptly while daily reports complete within their SLA.
Incorrect Answers
A:Configure an appropriate data gateway and scheduled refresh where needed. That action primarily handles Power BI scheduled refresh and gateway; the given requirement is Streaming versus batch latency choice. Streaming processes arriving records continually unlike scheduled batch.
B:Apply a justified real-time anomaly rule with quality checks and alert thresholds. That action primarily handles Streaming anomaly detection and false positives; the given requirement is Streaming versus batch latency choice. Streaming processes arriving records continually unlike scheduled batch.
C:Call a once-daily CSV read a continuous real-time stream. It does not implement Streaming versus batch latency choice. Running hourly batch for urgent alerts creates unacceptable delay.
Question 1136
A collection frequently filters by nested order status. Which choice BEST meets this requirement?
- Compare supported serverless and provisioned RU throughput for workload pattern.
- Issue a limited SAS or supported delegated access scoped by operation resource and expiry.
- Review supported indexing policy and partition-aware query patterns.
- Assume every unbounded cross-partition query costs the same as an ID point read.
Explanation
Correct Answer
C:Indexed document properties support queries while policy affects write/read cost.
Incorrect Answers
A:Compare supported serverless and provisioned RU throughput for workload pattern. That action primarily handles Cosmos DB serverless versus provisioned throughput; the given requirement is Cosmos DB NoSQL indexing and queries. Indexed document properties support queries while policy affects write/read cost.
B:Issue a limited SAS or supported delegated access scoped by operation resource and expiry. That action primarily handles Blob SAS delegated access; the given requirement is Cosmos DB NoSQL indexing and queries. Indexed document properties support queries while policy affects write/read cost.
D:Assume every unbounded cross-partition query costs the same as an ID point read. It does not implement Cosmos DB NoSQL indexing and queries. Excessive indexing or scans may inflate RU consumption.
Question 1137
A team must distribute requests by customer ID instead of a fixed country value. What is the MOST appropriate next validation or correction?
- Rehearse selected failover scenario and measure client routing.
- Check method scope expiry and recipient handling with test requests.
- Put the constant Global into every document partition key and assume perfect scale.
- Inspect key cardinality and request unit consumption per partition.
Explanation
Correct Answer
D:Inspect key cardinality and request unit consumption per partition. This tests the intended result: load tests avoid concentrating throughput on one logical partition.
Incorrect Answers
A:Rehearse selected failover scenario and measure client routing. It tests Cosmos DB multi-region distribution rather than the critical Cosmos DB logical partition key behavior. Partition keys determine distribution and point-read targeting.
B:Check method scope expiry and recipient handling with test requests. It tests Blob SAS delegated access rather than the critical Cosmos DB logical partition key behavior. Partition keys determine distribution and point-read targeting.
C:Put the constant Global into every document partition key and assume perfect scale. This unsafe shortcut leaves the Cosmos DB logical partition key problem unresolved. One hot key can throttle even when account capacity is increased.
Question 1138
A payment API must respond within 100 milliseconds while a loader moves terabytes nightly. Which choice BEST meets this requirement?
- Divide stored gigabytes by a request timeout and declare response latency solved.
- Evaluate response latency separately from aggregate processing throughput.
- Choose suitable storage redundancy and separate restore or backup strategy where needed.
- Run a scheduled batch ingestion with clear checkpoints and retries.
Explanation
Correct Answer
B:Latency measures time per operation; throughput measures completed work rate.
Incorrect Answers
A:Divide stored gigabytes by a request timeout and declare response latency solved. It does not implement Operational SLA latency versus throughput. High throughput alone cannot guarantee a low p99 request latency.
C:Choose suitable storage redundancy and separate restore or backup strategy where needed. That action primarily handles Storage resiliency and replicas; the given requirement is Operational SLA latency versus throughput. Latency measures time per operation; throughput measures completed work rate.
D:Run a scheduled batch ingestion with clear checkpoints and retries. That action primarily handles Batch scheduled processing; the given requirement is Operational SLA latency versus throughput. Latency measures time per operation; throughput measures completed work rate.
Question 1139
Financial source files contain malformed currency codes. After selecting the proper Batch ETL staging and cleansing approach, which observation gives the clearest evidence of success?
- The configured alert fires only for matching test conditions.
- Curated rows use consistent codes while malformed rows are quarantined.
- Approved device identities send telemetry while revoked devices are rejected.
- Queries scan fewer selected columns while data types remain valid.
Explanation
Correct Answer
B:Curated rows use consistent codes while malformed rows are quarantined. This is the directly observable result for Batch ETL staging and cleansing; eTL applies transformations before data reaches the analytical target.
Incorrect Answers
A:The configured alert fires only for matching test conditions. This supports Fabric Activator rule-triggered action, not confirmation of Batch ETL staging and cleansing. The relevant evidence is that curated rows use consistent codes while malformed rows are quarantined.
C:Approved device identities send telemetry while revoked devices are rejected. This supports Azure IoT Hub device messaging, not confirmation of Batch ETL staging and cleansing. The relevant evidence is that curated rows use consistent codes while malformed rows are quarantined.
D:Queries scan fewer selected columns while data types remain valid. This supports Parquet compressed columnar analytics, not confirmation of Batch ETL staging and cleansing. The relevant evidence is that curated rows use consistent codes while malformed rows are quarantined.
Question 1140
A business asks which role coordinates relational backup integrity and recovery. During testing, the team observes this specific failure: Assuming cloud-managed storage always eliminates backup planning risks unrecoverable mistakes. Which corrective check should be performed FIRST?
- Analyze execution plans indexes and database workload metrics.
- Perform a test restore and inspect retention and permissions.
- Store binary images as objects and maintain searchable metadata separately.
- Delegate all database recovery decisions to a report chart designer.
Explanation
Correct Answer
B:Perform a test restore and inspect retention and permissions. A successful correction should demonstrate that a controlled restoration meets stated RPO and RTO objectives.
Incorrect Answers
A:Analyze execution plans indexes and database workload metrics. It serves DBA performance tuning and indexing, not the diagnosed DBA backup and recovery responsibility problem. Assuming cloud-managed storage always eliminates backup planning risks unrecoverable mistakes.
C:Store binary images as objects and maintain searchable metadata separately. It serves Unstructured image binary files, not the diagnosed DBA backup and recovery responsibility problem. Assuming cloud-managed storage always eliminates backup planning risks unrecoverable mistakes.
D:Delegate all database recovery decisions to a report chart designer. This reinforces the error rather than addressing the DBA backup and recovery responsibility failure. DBAs manage database reliability security and recoverability within their responsibility boundary.