DP-203: Azure Data Engineer Associate Practice Exam

What the DP-203 exam covers

• Design and Implement Data Storage209 questions
• Develop Data Processing194 questions
• Secure, Monitor, and Optimize Data Storage and Processing194 questions

Free DP-203 sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 597.

1. Question 1Design and Implement Data Storage

   A retail analytics team stores daily point-of-sale transaction files in Azure Data Lake Storage Gen2. Each file is approximately 500 MB and is partitioned by date (year/month/day). Analysts report that queries filtering by a specific store location within a single day are slow because the entire daily file must be scanned. You need to improve query performance for location-based filters without changing the existing date-based folder hierarchy. What should you do?

   • AEnable hierarchical namespace on a new storage account and copy the data
   • BIncrease the Data Lake Storage Gen2 account tier from Standard to Premium
   • CRepartition the data within each daily folder by store location using additional subfolders and Parquet files with row group statisticsCorrect
   • DConvert the Parquet files to CSV format to reduce file complexity
   ✓ Correct answer: C

   Repartitioning each daily folder by store location into subfolders combined with Parquet row group statistics lets query engines like Synapse serverless SQL and Spark perform partition elimination (skipping folders for non-matching stores) and row-group pruning using min/max statistics, so only the relevant data is read. Because the date-based hierarchy is preserved and a new sub-level is added beneath it, existing date queries continue to work while location filters become selective. Parquet stores column statistics per row group, enabling predicate pushdown so the reader skips row groups that cannot match the store filter. This is the standard ADLS Gen2 optimization for selective columnar reads without re-architecting the lake.

   Why the other options are wrong

   • AEnabling hierarchical namespace on a new account does not change how data is physically laid out within each daily file; the entire file would still be scanned for a store filter, so the slow query pattern is unchanged.
   • BIncreasing the account from Standard to Premium tier only lowers per-operation latency and raises throughput for high-transaction workloads; it does not provide partition elimination or data skipping for a location predicate within a large file.
   • DConverting Parquet to CSV removes columnar storage, compression, and row group statistics, eliminating predicate pushdown and forcing full scans, which makes location-based queries slower, not faster.
2. Question 2Design and Implement Data Storage

   Contoso Ltd needs to store streaming telemetry data with a retention period of 7 days for real-time processing and also archive the data permanently in Azure Data Lake Storage Gen2. Which Azure service provides built-in capture capability to automatically archive streaming data to the data lake?

   • AAzure Queue Storage with lifecycle management
   • BAzure IoT Hub with message enrichment
   • CAzure Service Bus with auto-forwarding
   • DAzure Event Hubs with Capture featureCorrect
   ✓ Correct answer: D

   Azure Event Hubs with the Capture feature automatically archives streaming events to ADLS Gen2 (or Blob) in Avro at configurable time/size intervals while the hub still retains data for short-term real-time processing. Capture provides no-code, built-in archival of the stream to the lake. This directly meets both the short retention for real-time and permanent lake archival requirements.

   Why the other options are wrong

   • AAzure Queue Storage is a simple message queue without a built-in streaming-to-lake capture capability, so it cannot auto-archive telemetry to the data lake.
   • BAzure IoT Hub message enrichment augments message metadata but does not itself provide the built-in capture-to-ADLS archival the question requires (it relies on routing/Event Hubs Capture downstream).
   • CAzure Service Bus auto-forwarding chains queues/topics together; it has no native feature to automatically archive a stream to the data lake.
3. Question 3Design and Implement Data Storage

   What is the recommended file size range for Parquet files stored in Azure Data Lake Storage Gen2 for optimal query performance with Synapse Analytics?

   • A1 GB to 10 GB
   • B10 MB to 100 MB
   • C100 MB to 1 GBCorrect
   • D1 MB to 10 MB
   ✓ Correct answer: C

   The recommended Parquet file size for optimal Synapse query performance is roughly 100 MB to 1 GB. Files in this range amortize per-file open and metadata overhead while keeping row groups large enough for efficient columnar scans and parallelism. Too-small files create overhead and too-large files reduce parallelism, so the 100 MB-1 GB band is the standard guidance.

   Why the other options are wrong

   • A1 GB to 10 GB files are larger than recommended and can limit parallelism and increase memory pressure per file.
   • B10 MB to 100 MB skews toward small files, adding open and metadata overhead that hurts performance.
   • D1 MB to 10 MB are small files that cause significant overhead and are exactly what compaction tries to avoid.
4. Question 4Develop Data Processing

   You have an Azure Data Factory pipeline with activities A, B, and C. Activity C should run only if Activity A succeeds and Activity B fails. How should you configure the dependency conditions?

   • AA -> C (On Completion), B -> C (On Completion)
   • BA -> C (On Success), B -> C (On Failure)Correct
   • CUse an If Condition activity after both A and B
   • DA -> B -> C with conditional expressions
   ✓ Correct answer: B

   In Azure Data Factory, activity dependency conditions let you trigger a downstream activity based on each upstream activity's outcome, and multiple dependencies on the same activity are combined with logical AND. Configuring A -> C as On Success and B -> C as On Failure means Activity C runs only when A has succeeded and B has failed. A and B run independently, and C's two dependency conditions must both be satisfied for it to execute.

   Why the other options are wrong

   • AA -> C (On Completion), B -> C (On Completion) is incorrect because the On Completion condition fires regardless of whether the upstream activity succeeded or failed. Activity C would therefore run after both A and B finish in any state, failing to enforce that A must succeed and B must fail.
   • CUse an If Condition activity after both A and B is incorrect because an If Condition cannot reliably branch on the prior activities' success or failure states the way native dependency conditions can. It also forces extra activities and expression logic instead of using the built-in dependency framework that directly models On Success and On Failure outcomes.
   • DA -> B -> C with conditional expressions is incorrect because chaining the activities makes B depend on A and C depend on B, creating a sequential flow. The requirement needs A and B to execute independently, with C gated on A succeeding and B failing, which a linear chain cannot represent.
5. Question 5Develop Data ProcessingSelect all that apply

   You are implementing a data pipeline using Azure Data Factory. Which TWO trigger types support defining a schedule based on a recurring time interval? (Choose two.)

   • AUmbling window triggerCorrect
   • Bvent-based trigger (Blob storage)
   • Custom event trigger
   • DChedule triggerCorrect
   ✓ Correct answer: A, D

   Tumbling window trigger creates a recurring trigger that executes a pipeline on a regular interval (e.g., every hour, every day) with tumbling (non-overlapping) windows, allowing you to define schedules based on fixed recurring intervals. Schedule trigger provides a way to schedule pipelines using cron expressions or specific time intervals. Both support defining recurring time-based schedules, unlike event-based triggers which depend on resource events. Event-based trigger (Blob storage) is incorrect because event-based triggers fire when a specific event occurs (e.g., file created in blob storage), not on a recurring time interval schedule.

   Why the other options are wrong

   • BThe trigger depends on the event occurring, not on time.
   • CCustom event trigger is incorrect because custom event triggers are used to respond to custom events published to Event Grid, not for time-based scheduling. Custom events are external events, not recurring time intervals.
6. Question 6Secure, Monitor, and Optimize Data Storage and Processing

   You need to monitor Azure Data Factory pipeline failures and send email notifications to the operations team. Which approach should you use?

   • AConfigure Azure Monitor alerts on ADF pipeline failed metricsCorrect
   • Bcheck the ADF Monitor tab manually each morning
   • CWrite a custom application that polls the ADF REST API
   • DAdd a Web activity at the end of each pipeline to send emails
   ✓ Correct answer: A

   Azure Data Factory emits metrics such as failed pipeline runs to Azure Monitor, which can evaluate those metrics and fire alerts automatically. By creating an alert rule on the failed-pipeline metric and attaching an action group with an email receiver, the operations team is notified the moment failures occur without any custom polling code. This is the native, supported, and lowest-maintenance approach to failure notification.

   Why the other options are wrong

   • BChecking the ADF Monitor tab manually each morning is reactive and delayed, missing failures until someone happens to look rather than alerting in near real time.
   • CWriting a custom application that polls the ADF REST API reinvents functionality Azure Monitor already provides and adds maintenance burden and latency.
   • DAdding a Web activity at the end of each pipeline to send emails only fires on the paths that reach it and must be duplicated in every pipeline, making it brittle compared to centralized metric alerts.
7. Question 7Secure, Monitor, and Optimize Data Storage and ProcessingSelect all that apply

   You are optimizing the cost of an Azure Synapse Analytics solution. The dedicated SQL pool is used heavily during business hours (8 AM to 6 PM) but is idle overnight and on weekends. Which TWO strategies should you implement to reduce costs?

   • ADelete the SQL pool each night and recreate it each morning
   • Bschedule pause/resume using Azure Automation or Synapse REST APICorrect
   • CScale down to a lower DWU level during off-peak hours and scale up during peak hoursCorrect
   • DConfigure auto-pause to automatically pause the pool during idle periods
   • EConvert the dedicated SQL pool to a serverless SQL pool
   ✓ Correct answer: B, C

   To reduce cost for a dedicated SQL pool that is idle overnight and on weekends, schedule pause/resume using Azure Automation or the Synapse REST API (pausing stops compute billing), and scale down to a lower DWU during off-peak and up during peak hours. Pausing eliminates compute charges when unused, and scaling matches DWUs to demand. These two strategies are the standard cost optimizations.

   Why the other options are wrong

   • ADeleting and recreating the pool nightly risks data and configuration loss and is far more disruptive than pausing, which preserves data while stopping compute charges.
   • DDedicated SQL pools do not have an auto-pause-on-idle feature like serverless options; pausing must be scheduled/triggered, so 'configure auto-pause' is not an available mechanism.
   • EConverting to a serverless SQL pool is a re-architecture with different query semantics and storage, not a simple cost toggle for an existing dedicated pool workload.
8. Question 8Secure, Monitor, and Optimize Data Storage and Processing

   Which Azure Synapse feature allows administrators to restrict network access to the workspace using private endpoints?

   • AAzure Front Door
   • BSynapse Managed VNetCorrect
   • CNetwork Security Groups only
   • DAzure Firewall
   ✓ Correct answer: B

   The Synapse Managed VNet places the workspace's compute in an isolated, Microsoft-managed virtual network and enables managed private endpoints, which connect the workspace to data sources over private links so access does not traverse the public internet. Enabling the managed virtual network is the workspace setting that unlocks this private-endpoint connectivity model. It is therefore the Synapse feature that restricts network access using private endpoints.

   Why the other options are wrong

   • AAzure Front Door is a global HTTP/S load balancer and CDN for web applications, not a mechanism for private-endpoint network isolation of a Synapse workspace.
   • CNetwork Security Groups only filter subnet and NIC traffic by rules but do not provide the managed private-endpoint connectivity that isolates the workspace from the public network.
   • DAzure Firewall is a centralized network firewall for filtering and inspecting traffic, not the workspace feature that enables private-endpoint access to Synapse.
9. Question 9Secure, Monitor, and Optimize Data Storage and Processing

   A team is planning Data procedures for Secure Monitor and Optimize Data Storage and Processing. What should they prioritize?

   • AImplement role-based access control with least privilegeCorrect
   • BUse a single shared service account for the entire team
   • CGrant full administrator access to all team members
   • DDisable access controls for faster day-to-day workflows
   ✓ Correct answer: A

   Implementing role-based access control with least privilege is the recommended approach, because granting each identity only the permissions it needs minimizes the attack surface and limits the impact of any compromised account. Azure RBAC and data-plane ACLs let you scope access precisely to resources and actions. Least privilege is a core security principle for data platforms handling sensitive data.

   Why the other options are wrong

   • BUse a single shared service account for the entire team advocates a single shared service account, which destroys individual accountability and violates least-privilege principles.
   • CGrant full administrator access to all team members advocates granting everyone full admin access, which violates least privilege and dramatically increases the blast radius of any compromise.
   • DDisable access controls for faster day-to-day workflows advocates disabling access controls, which removes least-privilege protection and exposes data to unauthorized access.
10. Question 10Develop Data ProcessingSelect all that apply

    A consultant is reviewing the watermark for incremental load configuration at Contoso Ltd. Which two actions should be performed to optimize the implementation? (Choose two.)

    • ADisable watermark for incremental load monitoring
    • Bstream processing with Event Hubs
    • Cdata flow transformations
    • DAzure integration runtimeCorrect
    • Enotebook orchestrationCorrect
    ✓ Correct answer: D, E

    To optimize this watermark-based incremental-load implementation, the answer key selects the Azure integration runtime and notebook orchestration. The Azure integration runtime provides the managed compute and connectivity that move data for the incremental activities. Notebook orchestration coordinates the execution of the notebooks that compute and apply the watermark logic in the correct order. Together they address both the connectivity and the execution-coordination aspects of the solution.

    Why the other options are wrong

    • ADisabling watermark for incremental load monitoring removes visibility into the load's progress and correctness, which hinders optimization rather than helping it.
    • BStream processing with Event Hubs targets real-time event ingestion and is not one of the two components the answer key names for this batch incremental load.
    • CData flow transformations provide code-free reshaping but are not the components selected here, which center on the integration runtime and notebook orchestration.

DP-203 practice exam FAQ