DP-203 Exam Questions and Answers

Enable Transparent Data Encryption (TDE) with a customer-managed key in Azure Key Vault.

Transparent Data Encryption operates at the storage layer, encrypting data and log files at rest, and supports a customer-managed key held in Azure Key Vault — satisfying the stem's requirement that the dedicated SQL pool's data at rest be encrypted under organisational key control.

Deploy the workspace in a VNet with forced tunneling enabled and a firewall.

Deploying into a customer-managed VNet with forced tunnelling routes all cluster node egress through your firewall appliance, satisfying the outbound inspection requirement. Secure cluster connectivity (no public IP) is enabled alongside, so nodes hold no public addresses. This meets both constraints the security team specified.

Set ACLs on the directory with default ACLs for the service principal.

Default ACLs apply only to new child items created within the directory, not to the directory itself or to existing objects, so they cannot grant the service principal read and write access to the target directory directly. This option is tempting because default ACLs are designed to propagate permissions to future files and subdirectories, making them correct when the requirement is to control access for newly created content rather than the existing directory.

Queries classified as 'HeavyLoader' will not have a timeout.

A queryExecutionTimeoutSeconds value of zero disables the timeout entirely, so 'HeavyLoader' queries run until completion or manual cancellation. This satisfies the stem's scenario by confirming that the classifier imposes no execution time limit, unlike classifiers with a positive value that terminate queries after the specified seconds.

sys.dm_exec_requests

sys.dm_exec_requests shows currently executing requests in the serverless SQL pool, including state, command, and session ID.

Column-level security (CLS)

Column-level security uses GRANT and DENY statements on individual columns, restricting which roles can read specified fields within the table. This directly enforces the requirement that only users holding a certain role view particular columns.

Broadcast the smaller table (1 million rows) to all worker nodes.

Broadcasting the smaller table avoids shuffling the large table, significantly reducing data movement.

Increase the executor memory setting in the Spark configuration.

Executor out-of-memory errors arise when each executor's JVM heap cannot hold the partition data during transformations. Raising spark.executor.memory gives those executors more heap, directly relieving the constraint. Partition tuning or skew handling may follow, but increasing memory is the quickest first action.

Use coalesce() to reduce the number of partitions without a shuffle.

coalesce() merges partitions into the requested count without a full shuffle, so each task writes fewer, larger files. On a 100-million-row DataFrame this directly reduces the small-file problem while avoiding the network cost of repartitioning.

Use repartition() with a smaller number of partitions.

repartition() with a smaller partition count performs a full shuffle that redistributes rows evenly, so each output task writes one larger file. This reduces the number of small files, though it costs more than coalesce() because of the shuffle.

Use a Copy activity with pagination rules enabled in the source.

Copy activity pagination rules let the source iterate REST API pages automatically using properties such as NextPageUrl or absolute URLs, retrieving all 50,000 records without a ForEach loop. This satisfies the stem's requirement to traverse 1,000-record pages efficiently.

SELECT DeviceType, COUNT(*) FROM Input GROUP BY DeviceType, TumblingWindow(second, 10)

TumblingWindow(second, 10) partitions events into fixed, non-overlapping 10-second intervals, satisfying the stem's requirement to count per device type every 10 seconds. Grouping by DeviceType alongside the window produces one count per device type per interval, which sliding or session windows cannot guarantee.

Azure Databricks with Structured Streaming

Azure Databricks Structured Streaming natively supports Delta Lake sinks with idempotent writes and checkpointing, delivering exactly-once semantics when consuming Event Hubs. This satisfies the stem's exactly-once requirement, which plain Spark or Synapse streaming cannot guarantee without additional transactional handling.

Organize folders as /YYYY/MM/DD/DeviceID/ in ADLS Gen2 and use file naming that includes timestamp.

ADLS Gen2 hierarchical namespace supports true directory semantics, so /YYYY/MM/DD/DeviceID/ paths let partition pruning skip irrelevant folders during queries. Date-first ordering suits time-range filters, while DeviceID narrows per-device scans, and timestamped filenames preserve ingestion order within each partition.

Enable Change Data Capture (CDC) on the source table and use Azure Data Factory with a 1-minute tumbling window to copy changes into Synapse.

CDC captures only changed rows, and ADF can run frequently to meet latency target.

Store the logs in Azure Blob Storage with cool access tier.

Blob Storage cool tier is low-cost for infrequent access, suitable for logs.

Azure Data Lake Storage Gen2

ADLS Gen2 is a supported source.

Azure Blob Storage

Azure Blob Storage is a supported source.

You can query the external table using standard T-SQL SELECT statements.

Dedicated SQL pools expose external tables through the PolyBase engine, so standard T-SQL SELECT statements work against them exactly as with internal tables. The external table definition supplies the schema and the LOCATION pointing to the data in Azure Data Lake Storage, satisfying the requirement to query without loading.

Azure Event Hubs

Event Hubs ingests millions of telemetry events per second with low latency, buffering the stream for downstream consumers. This satisfies the real-time dashboard requirement while retaining data for long-term analytics, unlike batch-oriented stores such as Blob Storage or Azure SQL Database.