DEA-C01
Free DEA-C01 practice questions
10 questions from our DEA-C01 practice tests, each with the correct answer and a written explanation. Read them with the answers shown, or switch to quiz mode and answer first.
Question 1 of 10, Data Ingestion and Transformation
A company must move 600 TB of historical archives from an on-premises data center into Amazon S3 within a few weeks. Its 100 Mbps internet link would need more than a year to copy the data, but the company can have a high-bandwidth network connection to AWS provisioned through a partner in time. Which approach is MOST practical?
- AKeep the 100 Mbps link and run several AWS DataSync agents in parallel to multiply the throughput
- BKeep the 100 Mbps link and enable S3 Transfer Acceleration so CLI uploads use the nearest edge location
- CRead the archives with a custom producer and stream them into S3 through Amazon Data Firehose
- DProvision a 10 Gbps AWS Direct Connect connection and copy the archives with AWS DataSync tasks over it Correct answer
Answer and explanation: A 10 Gbps Direct Connect connection moves 600 TB in roughly a week at line rate, and DataSync handles parallel transfer, integrity verification and retries, so the deadline of a few weeks is met. Transfer Acceleration speeds up long-distance paths but cannot exceed the 100 Mbps local link. Extra DataSync agents still share the same 100 Mbps link, so the copy still takes over a year. Firehose is meant for streaming records, not for bulk migration of large archive files.
Question 2 of 10, Data Ingestion and Transformation
A data engineer needs to invoke a Lambda function that kicks off an ingestion job every weekday at 07:30 in the America/New_York time zone, adjusting automatically for daylight saving time. Which service should be used?
- AEventBridge Scheduler with a cron expression and a time zone Correct answer
- BAn S3 event notification on a trigger prefix that a cron script writes to each morning
- CAn Amazon SQS delay queue that a Lambda function polls and re-enqueues each day
- DAn AWS Glue crawler on a weekday schedule whose completion event triggers the function
Answer and explanation: EventBridge Scheduler supports cron schedules in a specified time zone, including daylight saving adjustments, and can invoke Lambda directly. An S3 event notification responds to object events and would still need a separate scheduler to write the trigger object. SQS delay queues delay messages by up to 15 minutes and cannot schedule recurring jobs by wall-clock time. Crawlers discover schemas, and their schedules are not a general way to invoke functions.
Question 3 of 10, Data Ingestion and Transformation
An AWS Glue ETL job runs every hour and reads new CSV files that land in an S3 prefix. The job currently reprocesses every file in the prefix on each run, which is slow and creates duplicate rows in the target. What is the MOST efficient way to process only files that arrived since the previous successful run?
- AAdd an S3 Lifecycle rule that expires files older than one hour from the source prefix after each run
- BRewrite the job as a Python shell job that lists objects and filters them by their LastModified timestamp
- CRun a Glue crawler before each job run so the Data Catalog table lists only the newly arrived partitions
- DEnable Glue job bookmarks and set transformation_ctx on the source and sink operations Correct answer
Answer and explanation: Glue job bookmarks persist state about which data was already processed, so subsequent runs pick up only new files; the transformation_ctx parameter is how Glue keys that state. Deleting source data with a lifecycle rule risks data loss and is not tied to job success. A crawler updates schema and partitions but does not stop the job from reading old files. Hand-built LastModified logic in a Python shell job is more code and loses Spark scale.
Question 4 of 10, Data Store Management
A company stores raw logs in S3. Logs are queried often during the first 30 days, rarely between 30 and 180 days, and must be kept for 7 years for compliance but are almost never retrieved. What is the MOST cost-effective approach?
- ALifecycle rules: Standard-IA after 30 days, Glacier Deep Archive after 180 days, expire at 7 years Correct answer
- BStore all logs in S3 One Zone-IA from the first day and transition them to Glacier Flexible Retrieval after 180 days
- CMove logs older than 30 days to Amazon EBS Cold HDD (sc1) volumes attached to an archive instance for 7 years
- DKeep all logs in S3 Standard with Intelligent-Tiering disabled and delete them with a scheduled script after 7 years
Answer and explanation: Lifecycle transitions match storage class to access pattern automatically: Standard-IA for infrequent access and Glacier Deep Archive for long-term, rarely retrieved retention, with expiration at 7 years. Keeping everything in Standard costs far more. One Zone-IA from day one adds retrieval fees during the frequent-access period and stores compliance data in a single Availability Zone, and Glacier Flexible Retrieval costs more than Deep Archive for data that is almost never read. EBS volumes are block storage and are much more expensive for archive.
Question 5 of 10, Data Store Management
Analysts need to join a Redshift dimension table with 200 TB of historical clickstream data stored as Parquet in S3 and cataloged in AWS Glue, without loading the historical data into the cluster. What should the data engineer use?
- AA Redshift COPY command that loads the S3 data into a temporary table at the start of each query
- BAn Athena federated query that joins the S3 data with the Redshift table through a Lambda connector
- CA Redshift zero-ETL integration that continuously replicates the S3 clickstream data into the cluster
- DRedshift Spectrum with an external schema that points to the Glue Data Catalog database Correct answer
Answer and explanation: Redshift Spectrum queries data in place in S3 through external tables defined in the Glue Data Catalog and can join it with local Redshift tables. Loading 200 TB per query with COPY is impractical. Athena federation runs the join in Athena, not in Redshift where the dimension lives, and adds a Lambda connector. Zero-ETL integrations replicate operational sources into Redshift, which loads the history into the cluster instead of querying it in place.
Question 6 of 10, Data Store Management
A data lake team needs ACID transactions, row-level updates and deletes for GDPR requests, schema evolution and time travel on large S3 datasets queried by Athena and Spark, while minimizing maintenance such as compaction. What should the data engineer use?
- AApache Iceberg tables in Amazon S3 Tables, which run automatic compaction Correct answer
- BAmazon DynamoDB tables with point-in-time recovery, exported to S3 for Athena and Spark
- CPlain Parquet files in S3 cataloged by Glue crawlers and rewritten by Spark jobs for deletes
- DCSV files in S3 with S3 Versioning enabled so earlier versions of each file can be restored
Answer and explanation: Apache Iceberg provides ACID transactions, row-level deletes, schema evolution and time travel, and S3 Tables stores Iceberg tables in table buckets with automatic compaction and snapshot management. Plain Parquet rewritten by Spark jobs has no transactions or time travel and needs custom maintenance. DynamoDB is not the analytic store for Athena and Spark at lake scale, and exports are point-in-time copies. S3 Versioning tracks object versions, not table-level transactions.
Question 7 of 10, Data Operations and Support
A business SLA requires that the sales table in the data lake be updated within 2 hours of midnight each day. The data engineer wants an alert whenever data is stale, regardless of which pipeline step failed. What is the MOST robust approach?
- ACreate EventBridge rules that send an SNS notification whenever any Glue job or crawler in the pipeline changes to the FAILED or TIMEOUT state
- BAlarm on the data lake bucket's daily BucketSizeBytes metric and alert the team whenever the bucket does not grow from one day to the next
- CPublish a custom CloudWatch freshness metric (last successful load time) at the end of the pipeline and alarm when it is not updated by the deadline, treating missing data as breaching Correct answer
- DCreate CloudWatch alarms on the orchestration state machine's ExecutionsFailed and ExecutionsTimedOut metrics and route both alarms to the on-call SNS topic so that every failed or timed-out run pages someone
Answer and explanation: A freshness metric emitted at the end of the pipeline, with an alarm that treats missing data as breaching, catches staleness from any cause, including runs that hang or never start. Alarms on Glue job state changes or on state machine failures and timeouts only fire for steps that actually run and fail, so a skipped schedule or a stuck upstream step goes unnoticed. Bucket size covers the whole bucket, not the freshness of one table.
Question 8 of 10, Data Operations and Support
A team runs Spark jobs on Amazon EMR and wants to define data quality constraints, such as uniqueness and value ranges, as code that runs inside the Spark job and produces metrics. Which open-source library is designed for this?
- AApache Zeppelin notebooks
- BApache Oozie SLA monitoring
- CDeequ (PyDeequ) Correct answer
- DApache Sqoop with --validate
Answer and explanation: Deequ, developed by AWS, runs on Spark to compute data quality metrics and verify constraints such as completeness, uniqueness and ranges. Sqoop transfers data between Hadoop and relational databases, and its validation only compares row counts after a transfer. Oozie schedules Hadoop workflows, and its SLA features track timing, not data constraints. Zeppelin is an interactive notebook, not a constraint library.
Question 9 of 10, Data Security and Governance
A table was registered in Lake Formation, and fine-grained grants were created, but users with broad IAM Glue and S3 permissions can still read all columns through Athena. What is the MOST likely cause?
- AThe Athena workgroup does not encrypt query results with SSE-KMS, so Lake Formation cannot vend temporary credentials for the query
- BIAMAllowedPrincipals still has Super permission on the table, so the default IAM-based access control keeps granting full access Correct answer
- CThe users assume roles without MFA, and Lake Formation evaluates fine-grained grants only for sessions that were authenticated with MFA
- DThe table is stored as Parquet, and Lake Formation column-level filtering is enforced only for tables stored as CSV or JSON files
Answer and explanation: When IAMAllowedPrincipals retains Super permission, Lake Formation defers to IAM-only access control for that resource, so fine-grained grants are not enforced; revoking it (or using hybrid access mode deliberately) makes Lake Formation permissions apply. Column filtering works for Parquet tables, Lake Formation does not depend on MFA, and result encryption has no effect on credential vending.
Question 10 of 10, Data Security and Governance
A company replicates SSE-KMS encrypted data lake objects from us-east-1 to eu-west-1 for disaster recovery and wants applications in either Region to decrypt the data using the same key ID and key material, managed as one key. Which approach supports this?
- AExport the key material and import it manually into another key
- BUse one single-Region key in us-east-1 and call it from eu-west-1
- CUse SSE-S3 for the replicated copies only
- DUse AWS KMS multi-Region keys, with a replica key in the destination Region specified in the replication configuration Correct answer
Answer and explanation: Multi-Region keys share key ID and key material across Regions, so data encrypted in one Region can be decrypted with the replica in another, and S3 Replication can re-encrypt with the replica key. Relying on a key in another Region adds cross-Region dependency and fails during a regional outage. SSE-S3 changes key control. Manually importing key material is complex and error-prone.