DEA-C01 question patterns

Use Glue (serverless Spark) with the Data Catalog and a scheduler or trigger.

No clusters to size or patch, unlike EMR on EC2.

Use Kinesis on-demand, DynamoDB on-demand or Redshift Serverless.

Capacity follows load, so no shard or capacity planning.

Firehose delivers buffered records to S3, Redshift, OpenSearch or Iceberg.

Fully managed delivery; near real time, not real time.

Register each consumer with enhanced fan-out on Kinesis Data Streams.

Each gets its own 2 MB/s per shard instead of sharing.

Keep records up to 365 days, or raw data in S3, and replay.

Firehose does not retain data.

Use a higher-cardinality key (add a suffix), split the shard or use on-demand.

A skewed key creates a hot shard.

Use MSK (with MSK Connect) instead of rewriting for Kinesis.

Kafka API compatibility with managed brokers.

Watch IteratorAgeMilliseconds; add shards, consumers or Lambda parallelization factor.

Rising iterator age means the consumer lags.

Enable ReportBatchItemFailures, bisect on error, retry limit and an on-failure destination.

Otherwise the batch retries until the record expires.

Use the zero-ETL integration.

Managed replication with no ETL to maintain.

Create an AppFlow flow with a schedule or event trigger.

Managed SaaS connector, no code.

Schedule a DataSync task.

Handles transfer, verification and filtering.

Create a Transfer Family SFTP server backed by S3.

Keeps partner tooling, no servers.

Use S3 event notifications or EventBridge to trigger Lambda or Step Functions.

Event-driven, no polling.

Route Glue or Lambda egress through a NAT gateway with an Elastic IP.

Gives a stable address to allowlist.

Put RDS Proxy in front of the database and cap Lambda concurrency.

Pools connections and protects the database.

Buffer in SQS and limit consumer concurrency.

Smooths bursts and absorbs 429s.

Enable job bookmarks.

They track what has been processed.

Filter on partition columns with push_down_predicate.

Only those partitions are listed and read.

Use groupFiles or compact to files of 128 MB to 1 GB.

Small files add listing and task overhead.

Convert CSV to Parquet with CTAS or Glue, partition by filter columns, compress.

Athena bills by data scanned and columnar skips columns.

Use EMR (task nodes on Spot, data in S3).

Glue lacks that control and engine choice.

State machine with Retry, Catch and the Glue .sync integration.

Serverless orchestration with native integrations.

Use SAM templates with sam build and sam deploy.

Short syntax for serverless resources on CloudFormation.

Use Flink for stateful stream processing.

Handles windows, joins and checkpointed state.

Use DynamoDB with a well-distributed partition key.

Designed for key-value access.

Use a Redshift warehouse in a star schema.

Columnar MPP built for analytics.

Split into many equal compressed files and COPY with a manifest.

COPY loads in parallel across slices.

Create an external schema over the Glue Catalog.

Keeps cold data in S3.

Create an external schema to the RDS or Aurora database.

Queries operational data without ETL.

Distribute both tables on the join column.

Avoids moving rows across nodes.

Run MSCK REPAIR, a crawler, or use partition projection.

The catalog must know about each partition.

Transition to colder classes, then expire; clean up noncurrent versions and multipart uploads.

Automates cost and retention.

Use S3 Intelligent-Tiering.

Moves objects automatically without retrieval fees.

Set a TTL attribute.

Deletions don't consume write capacity.

Add Evaluate Data Quality with DQDL rules.

Rules run in the pipeline and emit metrics.

Use DataBrew recipes and profile jobs.

Visual with 250+ transforms.

Use QuickSight with SPICE for speed.

Managed BI for business users.

Run Insights queries on the log group.

Fast ad hoc searching.

Export or deliver logs to S3 and query with Athena.

Pay per scan and no cluster.

Grant column-level or data-filter access, use LF-Tags.

Fine-grained access without copies.

Store the secret with rotation enabled.

Built-in rotation; Parameter Store has none.

Allow the role in the key policy and IAM, and in the bucket policy.

AWS managed keys can't be shared cross-account.

Add an S3 gateway endpoint with an endpoint policy.

Private route, free.

Create a datashare and authorise the consumer.

Live data, no ETL.

Macie findings, then LF-Tags to restrict columns.

Detection plus enforcement.

Deny actions outside approved Regions and limit replication.

Preventive control across accounts.

Create an event data store and query with SQL.

No S3 or Athena setup.

Use Config history and rules.

CloudTrail shows who called the API; Config shows state.

Enable data events on the trail.

They're off by default.