Question 20
Two Kinesis analytics applications need separately registered consumer identities and dedicated shard read throughput so neither reader slows the other. Which TWO measures meet both needs? Choose TWO.
- SQS standard queue fan-in
- Kafka consumer groups
- Kinesis enhanced fan-out
- Kinesis partition key design
- Separate registered stream consumers
Explanation
Correct Answers
C:Enhanced fan-out gives registered consumers dedicated throughput for a Kinesis Data Streams shard; in this scenario, dedicated consumer throughput reduces interference; each registered reader can process at an independent rate while tracking its own position.
E:Applications can register distinct Kinesis stream consumers to obtain their own enhanced fan-out paths; for the additional requirement of creating separately managed read identities for independent applications, register independent consumer applications and map each to its own processing logic and permissions.
Incorrect Answers
A:A standard SQS queue buffers messages from multiple producers for a pool of competing worker consumers; this can support decoupling many asynchronous producers from workers, but would not address the need for isolating high-throughput independent stream readers.
B:Kafka consumer groups divide topic partitions among members within a group while different groups read independently; this helps with scaling workers for one Kafka subscription, not isolating high-throughput independent stream readers.
D:A Kinesis partition key determines shard assignment and per-key ordering for records; it would target spreading heavy producer traffic across shards, while the question is about isolating high-throughput independent stream readers.
Question 21
Multiple analysts repeat the same joins and columns against S3-backed tables, but no materialized copy is needed. Which SQL object fits?
- Athena SQL view
- SQL PIVOT transformation
- Redshift SQL window function
- SQL GROUP BY aggregate
Explanation
Correct Answer
A:An Athena view stores a query definition for reuse over underlying tables; in this scenario, create an Athena view whose definition references governed source tables; the underlying files remain the source of records.
Incorrect Answers
B:Pivoting turns distinct category values into separate columns for cross-tab reports in supported SQL engines; it would target presenting monthly category totals as columns, while the question is about exposing a stable SQL projection over lake data.
C:Window functions calculate values across related rows while preserving row-level output; its benefit is computing a rolling average for each device; that is different from exposing a stable SQL projection over lake data.
D:GROUP BY collapses input rows into groups and computes aggregate measures per distinct key; the current need is for exposing a stable SQL projection over lake data, not producing daily totals per store.
Question 22
An ingestion Lambda scales too aggressively and overloads its fixed-capacity RDS endpoint. Which Lambda setting bounds it?
- Lambda memory/CPU setting
- Lambda asynchronous invocation destinations
- Provisioned concurrency
- Reserved concurrency
Explanation
Correct Answer
D:Reserved concurrency allocates a concurrency limit and reservation to a specific Lambda function; in this scenario, cap reserved concurrency at the safe parallelism and evaluate retry or queue backpressure for throttled incoming events.
Incorrect Answers
A:Lambda allocates more CPU capacity as the configured memory size increases; it would target speeding CPU-intensive parsing in a Lambda transform, while the question is about protecting a downstream database from too many simultaneous function calls.
B:Lambda asynchronous destinations can send execution success or failure records to supported targets; its benefit is recording failed asynchronous enrichments for repair; that is different from protecting a downstream database from too many simultaneous function calls.
C:Provisioned concurrency pre-initializes a configured number of Lambda execution environments to reduce cold-start delay; the current need is for protecting a downstream database from too many simultaneous function calls, not stabilizing low-latency synchronous API responses.
Question 23
Multiple teams need consistent column definitions and S3 paths for one data-lake dataset. Which metadata object should they reference?
- Glue Data Catalog table
- Data Catalog table version history
- Athena named database
- Glue partition metadata
Explanation
Correct Answer
A:A Data Catalog table describes a dataset's location, columns and serialization for compatible query engines; in this scenario, store the table schema and location centrally so query and ETL jobs resolve the same dataset description.
Incorrect Answers
B:Data Catalog versions can preserve prior table metadata definitions for certain schema-management workflows; the current need is for sharing an S3 table definition across Athena and Glue, not inspecting a prior table schema after a crawler update.
C:An Athena database is a logical namespace for tables in the underlying catalog; this can support organizing named datasets for a SQL workgroup, but would not address the need for sharing an S3 table definition across Athena and Glue.
D:Catalog partition definitions map partition values to locations so compatible engines can prune object sets; this helps with registering new date partitions before reading them, not sharing an S3 table definition across Athena and Glue.
Question 24
A central topic receives alerts from finance and marketing ETL jobs, but each team should receive only its own events. What configuration helps?
- EventBridge failed-event DLQ
- CloudWatch alarm evaluation periods
- SNS subscription filtering
- CloudWatch alarm composite
Explanation
Correct Answer
C:SNS subscription filter policies select which topic messages are delivered to specific subscribers; in this scenario, use message attributes and subscription filter policies so each team subscribes to only its authorized alert category.
Incorrect Answers
A:A rule target can use an SQS dead-letter queue to retain undelivered events after retry exhaustion; its benefit is capturing notifications that fail delivery to the target; that is different from routing alerts by data pipeline ownership.
B:Alarm evaluation periods govern how many datapoints must breach a threshold before the alarm changes state; the current need is for routing alerts by data pipeline ownership, not suppressing alerts for harmless single-sample spikes.
D:Composite alarms combine the states of other alarms using logical expressions; this helps with paging only when error rate and backlog are both unhealthy, not routing alerts by data pipeline ownership.
Question 25
A Python ETL loops through ten million dataframe rows to apply simple arithmetic and is CPU bound. Which code optimization should be attempted?
- Set membership lookup
- Vectorized dataframe operation
- Chunked iterator processing
- SQL predicate pushdown
Explanation
Correct Answer
B:Vectorized data operations process columns or batches without per-row interpreter overhead; in this scenario, use vectorized expressions or supported Spark column functions so execution can operate efficiently on batches.
Incorrect Answers
A:Hash-based set membership generally offers efficient repeated presence checks in memory; the current need is for reducing per-row Python transformation overhead, not accelerating millions of repeated key lookups.
C:Chunking processes large inputs in bounded pieces rather than loading an entire dataset into memory; this helps with reading multi-gigabyte files within tight memory limits, not reducing per-row Python transformation overhead.
D:Predicate pushdown filters eligible records near the source and reduces data moved for ETL; although useful for avoiding needless rows in a JDBC extraction, it leaves the need for reducing per-row Python transformation overhead unmet.
Question 26
A function transforms new objects but returns AccessDenied when calling GetObject, despite the bucket existing. What identity permission should be reviewed?
- Lambda execution IAM role
- Secrets Manager scheduled rotation
- Glue job service role
- STS AssumeRole for cross-account processing
Explanation
Correct Answer
A:Lambda assumes its execution role to call permitted AWS APIs during function invocation; in this scenario, grant the execution role scoped S3 read permissions and check relevant bucket and KMS key policies before rerunning.
Incorrect Answers
B:Secrets Manager rotation workflows can update supported credentials and associated secret versions; its benefit is rotating a relational password without embedding updates in code; that is different from granting Lambda controlled read access to one S3 prefix.
C:A Glue job runs with an IAM role that authorizes its data, catalog and logging operations; the current need is for granting Lambda controlled read access to one S3 prefix, not enabling Glue to access controlled catalog and S3 resources.
D:STS AssumeRole returns temporary credentials for an authorized role trust relationship; this can support allowing an ETL account to read a partner account's approved objects, but would not address the need for granting Lambda controlled read access to one S3 prefix.
Question 27
A data lake needs one ETL launch at 02:30 in a named time zone, including daylight-saving transitions. Which scheduler best fits?
- EventBridge Scheduler
- Glue job trigger schedule
- Amazon MWAA DAG schedule
- Step Functions with EventBridge rule
Explanation
Correct Answer
A:EventBridge Scheduler invokes supported targets on configured schedules with time-zone options; in this scenario, create a time-zone-aware schedule targeting the job or orchestration entry point, rather than encoding local offset calculations.
Incorrect Answers
B:Glue scheduled triggers start designated Glue jobs according to a configured schedule; it would target running a single Glue ETL job on a fixed cadence, while the question is about launching a daily transform at a specified local time.
C:An Airflow DAG schedule defines recurring execution of an orchestration graph in Amazon MWAA; its benefit is scheduling a dependency-aware Airflow workflow; that is different from launching a daily transform at a specified local time.
D:An EventBridge scheduled rule can initiate a Step Functions state machine execution; the current need is for launching a daily transform at a specified local time, not periodically invoking a multi-step serverless data workflow.
Question 28
A recommendation system must search conceptually similar product descriptions even when they use different words. What representation is needed?
- Bedrock Knowledge Bases ingestion
- Vector similarity metric
- Text embedding creation
- Embedding model version consistency
Explanation
Correct Answer
C:Embedding models encode text into numeric vectors that represent semantic features for comparison; in this scenario, generate consistent embeddings for documents and queries, then compare them under the model's distance metric.
Incorrect Answers
A:Bedrock Knowledge Bases can ingest supported sources, create embeddings and support retrieval workflows; this helps with preparing a managed document corpus for retrieval-augmented generation, not converting product descriptions to vectors for similarity retrieval.
B:Cosine similarity and other distance functions rank embeddings by geometric closeness under chosen semantics; although useful for selecting a compatible similarity function for embeddings, it leaves the need for converting product descriptions to vectors for similarity retrieval unmet.
D:Mixing embeddings from incompatible model versions can make distance comparisons unreliable; its benefit is preserving comparability after an embedding-model upgrade; that is different from converting product descriptions to vectors for similarity retrieval.
Question 29
A legacy database uses vendor-specific stored procedures and table definitions that must be assessed before moving to Aurora PostgreSQL. Which capability helps?
- AWS DMS Schema Conversion
- Dual-write migration validation
- Schema assessment report
- AWS DMS change data capture
Explanation
Correct Answer
A:AWS DMS Schema Conversion assesses and converts supported database schema objects between engines; in this scenario, run an assessment and conversion of supported schema constructs, then manually remediate unsupported procedural differences.
Incorrect Answers
B:Validating row counts, checksums and business invariants detects differences between migrated and original data; the current need is for migrating proprietary database objects to a different engine, not proving a migrated target is consistent before traffic moves.
C:A schema assessment highlights compatibility gaps and constructs requiring manual remediation during migration; this can support identifying unsupported conversion features early, but would not address the need for migrating proprietary database objects to a different engine.
D:DMS CDC transfers ongoing supported source changes after baseline replication to keep a target synchronized; this helps with minimizing database migration cutover downtime, not migrating proprietary database objects to a different engine.
Question 30
An application must initiate an existing AWS Glue ETL job from code and obtain its run identifier. Which direct AWS API operation fits?
- Redshift Data API execution
- Step Functions Glue run-job integration
- EMR step submission
- Athena StartQueryExecution
- Glue job start API
Explanation
Correct Answer
E:The Glue StartJobRun API launches a configured ETL job and returns a run identifier for tracking; in this scenario, start the Glue job through its supported service integration and track the run status before releasing dependent results.
Incorrect Answers
A:The Redshift Data API can submit SQL statements to compatible Redshift environments using authorized calls; this helps with invoking warehouse SQL without maintaining a persistent client connection, not starting Glue ETL from an application workflow.
B:A Step Functions Glue service integration can start a job and wait for its run to finish; although useful for coordinating downstream steps after a Glue job completes, it leaves the need for starting Glue ETL from an application workflow unmet.
C:EMR steps run specified data-processing tasks such as Spark jobs on a compatible cluster; it would target launching Spark processing on an existing EMR cluster, while the question is about starting Glue ETL from an application workflow.
D:Athena APIs can start SQL queries and return query execution IDs for polling results; its benefit is running a governed S3 SQL check from code; that is different from starting Glue ETL from an application workflow.