A. Correct.
Data discovery and classification labels sensitive columns and supports auditing and reporting scenarios.

B. Incorrect.
Round-robin distribution affects table storage and query performance, not sensitive data labeling.

C. Incorrect.
A NAT gateway controls outbound networking and does not classify data.

D. Incorrect.
DNS forwarding routes name-resolution requests and does not audit PII access.

A. Incorrect.
A single wildcard usually matches files in one folder level and does not recursively traverse nested folders.

B. Incorrect.
A single CSV file path does not read nested Parquet folders.

C. Correct.
Recursive wildcards can reference files in subfolders below the target path.

D. Incorrect.
A system view does not define an external file path.

A. Incorrect.
Bastion and public IPs are networking resources and are not prerequisites for external tables.

B. Correct.
External tables depend on an external data source and an external file format that describe location and file structure.

C. Incorrect.
Dashboards and workbooks visualize data but do not define external storage access.

D. Incorrect.
CDN and WAF resources are unrelated to Synapse external table DDL.

A. Incorrect.
A single-node cluster does not support many concurrent users well.

B. Incorrect.
A stopped job cluster cannot serve concurrent interactive users.

C. Incorrect.
A SQL serverless pool is a Synapse SQL feature, not a Databricks cluster type.

D. Correct.
High Concurrency clusters are designed for shared interactive workloads, and autoscaling improves utilization.

A. Incorrect.
VM restarts do not describe data movement lineage.

B. Correct.
Data movement and transformations from supported services such as Azure Data Factory can populate Purview lineage.

C. Incorrect.
DNS expiration is unrelated to data catalog lineage.

D. Incorrect.
CDN cache refresh is not a data pipeline lineage source.

A. Incorrect.
XML indexes are for XML data and are not the usual fact-table storage pattern.

B. Incorrect.
Large analytical tables typically benefit from columnstore storage.

C. Correct.
Clustered columnstore indexes are the standard choice for large analytical fact tables because they compress and scan columns efficiently.

D. Incorrect.
Spatial indexes support spatial queries, not general sales aggregates.

A. Incorrect.
Replicated tables are best for smaller dimensions, not very large facts.

B. Incorrect.
Putting all rows in one distribution creates skew and poor parallelism.

C. Incorrect.
DNS round robin is unrelated to Synapse table distribution.

D. Correct.
Hash distribution on a suitable high-quality join key can improve parallelism and reduce data movement.

A. Correct.
sys.dm_pdw_nodes_db_partition_stats can help analyze row distribution across Synapse distributions.

B. Incorrect.
sys.server_principals stores security principals and does not show table skew.

C. Incorrect.
Database mail views are unrelated to Synapse distribution skew.

D. Incorrect.
Connection DMVs do not show row counts by distribution.

A. Incorrect.
CSV is row-oriented and less efficient for column pruning and compression.

B. Correct.
Parquet is columnar, compressed, and efficient for analytical queries that skip columns.

C. Incorrect.
JSON is flexible but usually larger and less efficient for analytical scans.

D. Incorrect.
Plain text lacks schema and columnar optimization.

A. Incorrect.
Owner at subscription is far broader than needed.

B. Incorrect.
Reader does not grant blob write access.

C. Correct.
Storage Blob Data Contributor at the container scope grants data read/write/list permissions without broad management access.

D. Incorrect.
Key operator permissions at tenant root do not match least-privilege data access.