AWS Data Analytics Specialty (DAS-C01) Practice Exam — 65 Questions

A) Kinesis Data Firehose B) Kinesis Data Streams C) SQS FIFO Queue D) Amazon MSK

Answer: B

Explanation: Kinesis Data Streams provides partition-key-based ordering, configurable data retention up to 7 days (default 24 hours), and real-time replay capability. Firehose does not support replay, and SQS FIFO has throughput limits.

A) Increase the number of shards B) Enable Enhanced Fan-Out C) Switch to Provisioned capacity mode D) Reduce GetRecords API call frequency

Answer: B

Explanation: Enhanced Fan-Out provides each registered consumer with a dedicated 2 MB/s throughput per shard. Standard GetRecords shares the 2 MB/s per shard across all consumers, while Enhanced Fan-Out gives each consumer its own dedicated pipe.

A) Lambda transformation + prefix expressions B) Format Conversion + Dynamic Partitioning C) Integrate with Glue ETL D) Use S3 Object Lambda

Answer: B

Explanation: Kinesis Data Firehose Format Conversion transforms JSON to Parquet/ORC using the Glue Data Catalog schema. Dynamic Partitioning extracts values from records using jq expressions or inline parsing to build dynamic S3 prefixes. Both features are configured natively in Firehose.

A) Use AWS Glue ETL to periodically copy full tables B) Use AWS DMS with a continuous replication task C) Use Kinesis Data Streams + Lambda D) Initial load with Snowball, then Direct Connect

Answer: B

Explanation: AWS DMS reads transaction logs from the source database to implement CDC. After an initial full load, the ongoing replication task continuously synchronizes changes. For heterogeneous migration (Oracle to Redshift), use SCT (Schema Conversion Tool) alongside DMS to convert the schema.

A) Build an AWS Direct Connect dedicated line B) Order multiple Snowball Edge Storage Optimized devices C) Use AWS Snowmobile D) Enable S3 Transfer Acceleration

Answer: C

Explanation: For petabyte-scale (100 PB+) data transfers, Snowmobile is the appropriate choice. Each Snowmobile can hold up to 100 PB. Internet or Direct Connect transfers at this scale would take too long and cost too much. Snowball Edge holds up to 80 TB per device, requiring many devices for petabyte-scale.

A) Kinesis Data Streams — better scalability B) Amazon MSK — Kafka compatibility and long retention support C) Kinesis Data Firehose — fully managed D) SQS — longer message retention

Answer: B

Explanation: Amazon MSK is a fully managed Apache Kafka service, enabling existing Kafka client code to work without modification. Retention can be set to unlimited. Kinesis supports up to 365 days retention but is not compatible with the Kafka API.

A) Double the number of shards B) Add a random prefix to the partition key for uniform shard distribution C) Enable Enhanced Fan-Out D) Disable KPL aggregation

Answer: B

Explanation: Hot shard problems (write concentration on specific partition keys) are resolved by distributing the partition key. Prepending a random prefix (e.g., a number in range 0 to N) distributes records across multiple shards. On the read side, you must fan-out across all shards to reconstruct the full dataset.

A) Tumbling Window B) Sliding Window C) Session Window D) Global Window

Answer: B

Explanation: A Sliding Window has a fixed size that moves forward at a defined interval. For example, a 30-minute window sliding every 1 minute always evaluates the most recent 30 minutes of data. Tumbling Windows are non-overlapping fixed intervals.

A) Single partition: year/month/day B) Composite partition: year/month/day/region C) No partitioning, use Athena Partition Projection D) Store all data under a single prefix with compression

Answer: B

Explanation: Aligning the partitioning scheme with query patterns minimizes data scanned by Athena. A year/month/day/region hierarchy enables partition pruning for both filter conditions simultaneously, maximizing performance.

A) Restrict access via S3 bucket policy on specific prefixes B) Apply Glue catalog resource-based policies at the table level C) Configure Lake Formation column-level security D) Set query filters per Athena workgroup

Answer: C

Explanation: Lake Formation provides fine-grained access control at the table, column, and row level. Column-level security restricts which columns are visible to specific IAM roles or users. S3 bucket policies only control at the file level and cannot enforce column-level restrictions.

A) Both tables: EVEN distribution B) Both tables: ALL distribution C) Large fact table: KEY distribution (join key); small dimension table: ALL distribution D) Both tables: AUTO distribution

Answer: C

Explanation: Using KEY distribution on both fact tables with the same join key means data is co-located on the same nodes, enabling local joins without network redistribution. Small dimension tables with ALL distribution are broadcast to every node, avoiding broadcast join overhead for large-to-small joins.

A) Use a Scan with FilterExpression B) Use a Query with KeyConditionExpression specifying UserID and an OrderDate range C) Create a GSI and query it D) Use DynamoDB Streams + Lambda to maintain an index

Answer: B

Explanation: Using a Query with the table's primary key (UserID as partition key, OrderDate as sort key) is the most efficient access pattern. Set KeyConditionExpression to UserID = :uid AND OrderDate >= :date. Scans read the entire table and are highly inefficient.

A) To independently scale compute and storage B) For maximum query performance through in-memory caching C) To automatically parallelize queries D) To directly access data in S3

Answer: A

Explanation: Redshift RA3 nodes separate compute and storage, allowing each to scale independently. Frequently accessed data is cached on local NVMe SSD, while the rest resides in Redshift Managed Storage (RMS) backed by S3. You can grow storage without increasing compute costs.

A) Keep all data in Hot storage B) Transition data: Hot → UltraWarm → Cold storage C) Export data older than 30 days to S3 D) Use Index State Management to delete data after 30 days

Answer: B

Explanation: OpenSearch Service provides Hot (fast NVMe SSD), UltraWarm (S3-backed, lower cost), and Cold (lowest cost) storage tiers. Use Index State Management (ISM) to define automated policies that move indices through these tiers based on age, balancing performance and cost.

A) Automatic file compression reduces storage costs B) ACID transaction support and automatic data compaction C) Real-time streaming data ingestion D) Automatic column-level encryption

Answer: B

Explanation: Lake Formation Governed Tables provide ACID transactions (atomicity, consistency, isolation, durability) on S3 data, ensuring data consistency during concurrent reads and writes. Automatic compaction resolves the small files problem. Row-level security is also supported.

A) Store S3 data in CSV format B) Store S3 data in Parquet format with partitioning C) Maximize the number of Redshift cluster nodes D) Limit each Spectrum slice to one file

Answer: B

Explanation: Redshift Spectrum achieves best performance with columnar formats like Parquet or ORC. Column pruning (reading only required columns) combined with partition pruning (reading only required partitions) drastically reduces the data scanned. CSV requires reading entire files.

A) Use Spot for master, core, and task nodes B) Use On-Demand for master and core nodes; Spot only for task nodes C) Use On-Demand for master; Spot for core and task nodes D) Use Reserved Instances for all nodes

Answer: B

Explanation: The master node manages the cluster, and core nodes store HDFS data. Using On-Demand for both master and core nodes ensures stability and data durability. Task nodes (additional compute only, no HDFS storage) can safely use Spot. Spot interruption of task nodes does not cause data loss.

A) DynamicFrame has no schema and handles all data types B) DynamicFrame tolerates schema inconsistencies (choice type) and provides AWS-specific transforms like relationalize C) DataFrame is always faster so DynamicFrame should be avoided D) DynamicFrame supports only structured streaming

Answer: B

Explanation: AWS Glue DynamicFrame represents columns with mixed types as a "choice" type. It provides AWS-specific transformations such as resolveChoice() and relationalize(). For performance-critical paths, you can convert DynamicFrame to DataFrame, process it, and convert back.

A) Schedule the Glue Crawler to run every minute B) Use Athena MSCK REPAIR TABLE or ADD PARTITION C) Use Lake Formation blueprints to auto-manage partitions D) Use an S3 event notification to trigger Lambda that updates Glue Catalog partitions

Answer: D

Explanation: An S3 ObjectCreated event triggers a Lambda function that calls glue:BatchCreatePartition to add only the new partition to the Glue Catalog. This is the most efficient approach. Athena MSCK REPAIR TABLE also works but becomes slower as the partition count grows.

A) EMR Serverless maintains a permanent cluster; EMR on EC2 uses ephemeral clusters B) EMR Serverless automatically scales without cluster provisioning and has no idle costs C) EMR Serverless only supports Hive, not Spark D) EMR on EC2 is always more cost-effective

Answer: B

Explanation: EMR Serverless automatically provisions resources when a job is submitted without you managing any cluster. There are no charges when no jobs are running. It supports Spark, Hive, and other frameworks. It is ideal for intermittent batch workloads.

A) Large-scale distributed Spark ETL processing B) No-code visual data preparation and cleaning C) Real-time streaming data transformation D) Data catalog metadata management

Answer: B

Explanation: AWS Glue DataBrew provides a visual, code-free interface for exploring, cleaning, and normalizing data. It offers 250+ pre-built transformations and supports data quality rule definitions and profiling. It is designed for data analysts and data scientists who prefer visual tooling.

A) Poll EMR status with a Lambda function and restart on failure B) Add a Retry block to the Step Functions state definition C) Detect EMR failure with a CloudWatch Alarm and send SNS notification D) Wrap the EMR job in a Glue Workflow

Answer: B

Explanation: Adding a Retry block to each task state in Step Functions enables automatic retries for specified error types. You can configure the number of retries, interval, and backoff rate. A Catch block handles the final failure case with an alternative path.

A) Reduce the number of Glue workers B) Use coalesce() or repartition() to control the output partition count before writing C) Use S3 Lifecycle policies to automatically delete small files D) Merge files using Kinesis Firehose

Answer: B

Explanation: In Spark, coalesce(N) reduces the number of partitions without a shuffle (may produce uneven sizes), while repartition(N) redistributes evenly (incurs shuffle). Using either before a write operation controls the number of output files. coalesce is generally preferred for merging small files.

A) Increase Lambda function memory B) Increase the Parallelization Factor per shard C) Decrease the batch size D) Increase Lambda reserved concurrency

Answer: B

Explanation: The Parallelization Factor (1–10) for the Kinesis-Lambda event source mapping allows multiple concurrent Lambda invocations per shard. The default value of 1 means one concurrent Lambda execution per shard. Setting it to 10 enables up to 10 parallel Lambda invocations per shard simultaneously.

A) Athena Query Result Reuse B) Athena Federated Query C) CTAS to save results D) Athena workgroup query queuing

Answer: A

Explanation: Athena Query Result Reuse reuses previous query results for identical queries within a configurable period (up to 7 days). No scan charges are incurred for reused results. This is highly effective for repeated queries on static data.

A) Run MSCK REPAIR TABLE every hour B) Configure Athena Partition Projection C) Run ALTER TABLE ADD PARTITION every hour D) Use Lake Formation for automatic partition management

Answer: B

Explanation: Athena Partition Projection computes partition values dynamically from rules defined in the table properties, without storing partition metadata in the Glue Catalog. It is especially effective for regular patterns (dates, numeric ranges) and eliminates partition registration overhead.

A) Both tables use KEY distribution for efficient co-located joins B) One table uses ALL distribution so no data redistribution is needed for the join C) EVEN distribution requires data redistribution D) No distribution key, broadcast join not possible

Answer: B

Explanation: DS_DIST_ALL_NONE means one table in the join uses ALL distribution (copied to every node), so no redistribution is needed. This is an efficient join pattern. DS_DIST_ALL_INNER indicates redistribution cost is involved.

A) Direct Query mode connected to Redshift B) Import data into SPICE with a scheduled refresh C) Direct query of S3 data via Athena D) Use QuickSight Paginated Reports

Answer: B

Explanation: SPICE (Super-fast Parallel In-memory Calculation Engine) is QuickSight's in-memory storage that enables ultra-fast queries and visualization even for hundreds of millions of rows. Scheduled refresh keeps SPICE data current. Direct Query mode is real-time but may be slow for very large datasets.

A) Add nodes to the cluster B) Enable Short Query Acceleration (SQA) C) Assign equal priority to all queries D) Schedule long-running queries for nighttime

Answer: B

Explanation: Redshift Short Query Acceleration (SQA) uses machine learning to predict short-running queries and run them in a dedicated priority queue. It requires no separate WLM queue configuration and automatically optimizes environments where short and long queries coexist.

A) Create a separate dataset per sales representative B) Apply a QuickSight RLS (Row-Level Security) rule to the dataset C) Restrict S3 data access via IAM policies D) Use Athena views to filter data per user

Answer: B

Explanation: QuickSight RLS attaches a rules file (CSV or another dataset) mapping users/groups to filter values to a dataset. Data is automatically filtered based on the logged-in user. A single dataset serves all representatives, each seeing only their own region's data.

A) Join S3 data lake with RDS, DynamoDB, and other heterogeneous sources in a single SQL statement B) Query Redshift and S3 data together C) Process S3 buckets across multiple AWS accounts in a single query D) Analyze real-time streaming data with SQL

Answer: A

Explanation: Athena Federated Query uses Lambda-based data source connectors to run SQL on data sources beyond S3 (RDS, DynamoDB, ElastiCache, CloudWatch, Redis, etc.). Multiple heterogeneous sources can be joined in a single query, enabling integrated analytical workloads without data movement.

A) Reduce the Refresh Interval to 1 second B) Use the Bulk API and increase the Refresh Interval C) Call the Index API for each document individually D) Maximize the number of shards

Answer: B

Explanation: The Bulk API batches multiple documents per request, reducing network overhead. Increasing the Refresh Interval (e.g., from the default 1 second to 30 seconds or more) reduces segment creation frequency and significantly improves indexing throughput. Call a manual refresh after the bulk load completes.

A) S3 bucket policies alone provide fine-grained data access control B) Lake Formation data permissions for table/column/row-level access control C) IAM policies to control Glue Catalog access D) S3 Access Points to control access by prefix

Answer: B

Explanation: Lake Formation integrates with the Glue Data Catalog to provide logical table, column, and row-level fine-grained access control. Permissions are based on logical data structure rather than physical S3 file paths. S3 bucket policies only control at the file/prefix level and cannot enforce column or row-level restrictions.

A) Enable S3 server-side encryption (SSE-S3) B) Enable Server-Side Encryption (SSE) on the Kinesis stream — select the aws/kinesis KMS key C) Create a customer-managed KMS key (CMK) and attach it to Kinesis D) Use client-side encryption before sending data

Answer: B

Explanation: Kinesis Data Streams supports server-side encryption (SSE). Selecting the AWS-managed key (aws/kinesis) encrypts all stream data with KMS without additional configuration. A customer-managed key (CMK) provides more granular key policy control but requires additional setup.

A) Assign a public IP to the Redshift cluster B) Route S3 traffic through a NAT Gateway C) Configure a VPC endpoint (gateway endpoint) for S3 D) Use Direct Connect for AWS network connectivity

Answer: C

Explanation: An S3 gateway VPC endpoint routes Redshift COPY/UNLOAD operations through the AWS private network. S3 traffic does not traverse the internet gateway or NAT Gateway, improving security and eliminating data transfer costs.

A) It is automatically handled by S3 server-side encryption B) Enable metadata encryption in Glue Security Configuration C) Directly attach a KMS key to Glue Catalog tables D) Use Lake Formation to control metadata access

Answer: B

Explanation: AWS Glue Security Configuration enables encryption of Glue Data Catalog metadata with a KMS key. ETL job data encryption (at rest, in transit) and job bookmark encryption are also configured in Security Configuration.

A) Maintain a separate S3 bucket copy with PII removed for the analytics team B) Lake Formation column-level security + S3 Object Lambda for dynamic masking C) Deny full bucket access in the analytics team IAM role, allow only specific prefixes D) Use Glue ETL to generate a daily PII-free dataset

Answer: B

Explanation: Lake Formation column-level security blocks PII columns. S3 Object Lambda can dynamically mask PII at read time using a Lambda function. This approach avoids duplicating data and allows centralized management of masking logic.

A) Use CloudTrail to record Redshift API calls B) Enable Redshift Audit Logging to S3 C) Use VPC Flow Logs to record Redshift network traffic D) Use CloudWatch Logs to record Redshift connections

Answer: B

Explanation: Redshift Audit Logging writes connection logs, user activity logs, and user logs to S3. The user activity log records every executed SQL query. This provides complete Who/What/When information needed for compliance auditing. CloudTrail captures only Redshift API calls.

A) Kinesis Firehose → S3 → Athena (satisfies all requirements) B) Kinesis Data Streams → Lambda (real-time fraud detection) + Firehose → S3 (batch) + Glue + Redshift (reports) C) MSK → Spark Streaming → DynamoDB D) SQS → Lambda → RDS → QuickSight

Answer: B

Explanation: Kinesis Data Streams feeds Lambda for sub-100ms fraud detection and simultaneously routes to Firehose for S3 storage. Glue ETL processes S3 data and loads it into Redshift for daily reports. S3 Intelligent-Tiering manages 3-year retention cost-effectively.

A) Enable S3 server-side encryption (KMS) B) Enforce in-transit encryption (SSL/TLS) for Redshift C) Distribute data via CloudFront D) Store data in a public S3 bucket E) Use Kinesis streams without encryption

Answer: A, B

Explanation: HIPAA compliance requires encryption at rest (S3 SSE-KMS) and in transit (SSL/TLS). VPC deployment, access logging, and audit trails are also required. Options D and E violate HIPAA requirements, and CloudFront is not directly related to HIPAA data protection.

A) Copy data with S3 DistCp, regenerate metadata with Glue Crawler B) Physical data transfer with Snowball family, schema conversion with SCT C) Copy data with AWS DataSync, migrate Hive Metastore by importing into Glue Catalog D) Use DMS for the entire migration

Answer: C

Explanation: AWS DataSync efficiently transfers large datasets from on-premises HDFS to S3 with parallel transfer and automatic integrity verification. The Glue Data Catalog is compatible with the Hive Metastore and supports importing metadata. EMR can run existing Hive queries against the Glue Catalog.

A) Minimize file sizes to increase file count B) Optimize file sizes to 128 MB–1 GB to maintain splittable format C) Disable compression to speed up file reads D) Increase partitioning granularity with smaller partitions

Answer: B

Explanation: Parquet files are splittable by default. Maintaining the optimal file size (128 MB–1 GB) ensures maximum parallelism during Athena scans. Too-small files increase overhead; too-large files reduce parallelism. Combine with Snappy or ZSTD compression for best results.

A) Track Glue job execution history B) Track already-processed data to implement incremental processing C) Optimize Glue job costs D) Save data quality checkpoints

Answer: B

Explanation: Glue Job Bookmarks track which data was processed in previous runs. Subsequent runs process only newly added data, preventing duplicate processing. They use S3 file modification times and names to determine processing status. This enables efficient incremental ETL pipelines without reprocessing all data.

A) Load all reference data into Flink memory B) Use Flink Async I/O to asynchronously query the external database C) Send reference data as a Kinesis stream and join D) Enrich with a Lambda function and re-publish to Kinesis

Answer: B

Explanation: Flink Async I/O processes external database lookups (Redis, DynamoDB, etc.) asynchronously, minimizing processing latency. Synchronous lookups would block on external system responses and severely reduce throughput. Async I/O processes multiple outstanding requests concurrently. Consider local caching with RocksDB State Backend for further optimization.

A) Create standard QuickSight user accounts for each external user B) Use the QuickSight embedded URL API with anonymous embedding or Reader sessions C) Share public dashboard links D) Export dashboard images to S3 and display on the web

Answer: B

Explanation: The QuickSight embedding API generates time-limited URLs for external users to view dashboards without a QuickSight account. Anonymous embedding (unauthenticated access) or Reader sessions (per-session pricing) are available. The embedded dashboard is rendered in an iframe within the web application.

A) All records are deleted B) Failed records are stored in a separate S3 prefix (processing-failed prefix) C) Firehose automatically retries until success D) Failed records are returned to the source

Answer: B

Explanation: Records that fail Lambda transformation in Kinesis Data Firehose are written to the processing-failed S3 prefix. Successfully transformed records go to the configured destination. Failed records are preserved separately for later reprocessing or error analysis.

A) Copy each account's S3 data to the central account B) Use Lake Formation cross-account data sharing C) Configure S3 cross-account replication D) Use AWS Organizations SCP to consolidate data access

Answer: B

Explanation: Lake Formation supports cross-account data sharing. Data owner accounts grant Lake Formation data permissions to IAM roles/users in the central account. Athena or Redshift Spectrum in the central account can query other accounts' data catalogs. This achieves centralized governance without data movement.

A) Automatically converts Hive queries to Spark B) Supports UPSERT/DELETE and incremental processing on S3 C) Automatically scales the EMR cluster D) Automatically syncs data between HDFS and S3

Answer: B

Explanation: Apache Hudi enables UPSERT (insert/update), DELETE operations on data lake storage like S3, overcoming the immutability limitation of traditional data lakes. It provides Copy-on-Write and Merge-on-Read table types, and supports incremental queries to efficiently process only changed data.

A) Resolves out-of-memory issues in Glue ETL jobs B) Provides materialized views that replicate and combine data from multiple sources in near-real-time C) Fixes schema detection errors in Glue Crawlers D) Automatically caches Athena query results

Answer: B

Explanation: AWS Glue Elastic Views automatically replicates and combines data from source databases (DynamoDB, Aurora, RDS) to targets (OpenSearch, S3, Redshift) maintaining materialized views. A SQL-based view definition enables data integration without complex ETL pipelines.

A) Automatically encrypts all S3 operations B) Automatically monitors access patterns and moves data to the most cost-effective storage tier C) Automatically creates backups of S3 data D) Automates global data replication

Answer: B

Explanation: S3 Intelligent-Tiering monitors access frequency and automatically moves data: frequently accessed to Frequent Access tier, data not accessed for 30+ days to Infrequent Access, and 90+ days to Archive Instant Access. It is ideal for data lakes with unpredictable access patterns. Only a monitoring fee is added.

A) Terminate unnecessary database connections B) Reclaim storage from deleted/updated rows and re-sort data by sort key order C) Clear the Redshift cluster cache D) Update table statistics

Answer: B

Explanation: Redshift VACUUM physically removes rows that were only soft-deleted/updated and re-sorts data by the sort key. Regular VACUUM reclaims storage and maintains query performance. ANALYZE updates table statistics. Redshift Serverless and recent versions support automatic VACUUM.

A) Creating a new consumer group automatically starts from the beginning B) Use GetShardIterator API with AT_TIMESTAMP to start reading from a specific time C) Reprocess via S3 through Firehose D) Copy retention data to a new stream

Answer: B

Explanation: Setting ShardIteratorType to AT_TIMESTAMP in the Kinesis GetShardIterator API and specifying the desired timestamp lets you read data starting from that point. When using KCL, set the initial position to a timestamp.

A) Single metadata registry that shares schemas across multiple analytics engines, ensuring consistency B) Faster data processing speed C) Reduced S3 storage costs D) Automatic data quality validation

Answer: A

Explanation: The Glue Data Catalog is the central metadata store in AWS. Athena, Redshift Spectrum, and EMR all reference the same Glue Catalog, so table definitions, partition information, and schemas are managed in one place. Schema changes are automatically reflected across all analytics engines.

A) CloudWatch Alarms to monitor S3 file sizes B) AWS Glue Data Quality rules for automated validation C) Run manual data validation scripts with Lambda D) Send manual review notifications to the data team via SNS

Answer: B

Explanation: AWS Glue Data Quality defines quality rules (completeness, uniqueness, referential integrity, etc.) on datasets and automatically validates them during ETL jobs. It provides quality scores and rule pass/fail results. Integration with CloudWatch enables alerts when quality degrades.

A) Always cheaper than EMR on EC2 B) Run Spark jobs on existing Kubernetes infrastructure, integrating container-based workloads C) Configure HDFS storage on the EKS cluster D) GPU-based deep learning only

Answer: B

Explanation: EMR on EKS runs Apache Spark on an existing Amazon EKS cluster. Organizations already using Kubernetes can consolidate data processing workloads without managing separate EMR clusters. Benefits include container isolation, resource sharing, and integration with existing Kubernetes tooling (Helm, Argo Workflows, etc.).

A) Automatically adds cluster nodes for permanent expansion B) Automatically provisions additional cluster capacity during peak concurrency to maintain SLAs C) Automatically upgrades the Redshift cluster D) Automatically schedules overnight batch processing

Answer: B

Explanation: Redshift Concurrency Scaling automatically starts additional read clusters when concurrent queries saturate the main cluster. The first 24 hours per day are free; subsequent usage is billed per second. Users experience consistent, low-latency query performance without manual intervention.

A) No, DMS automatically handles schema conversion B) Yes, Aurora MySQL and Redshift are different engines; SCT is needed for schema conversion C) No, Aurora and Redshift are both AWS services so conversion is unnecessary D) SCT is only required for on-premises databases

Answer: B

Explanation: DMS handles data movement; SCT handles schema (DDL, procedures, functions, etc.) conversion. A heterogeneous migration from Aurora MySQL (OLTP) to Redshift (OLAP) requires data type and table structure conversion. Run SCT first to generate Redshift-compatible DDL, then use DMS to replicate the data.

A) Create a separate AWS account per team B) Use Athena Workgroups to isolate queries, track costs, and control data usage per team C) Use CloudWatch cost alarms to monitor budgets per team D) Use IAM tags to track query costs

Answer: B

Explanation: Athena Workgroups isolate queries by team or project and separately track query costs and data scanned per workgroup. You can set per-query data scan limits, specify result storage locations, and integrate with CloudWatch metrics. Combined with cost allocation tags, this enables team-level cost chargebacks.

A) Directly train and deploy machine learning models B) Detect anomalies, generate forecasts, and auto-narratives from time-series data C) Display SageMaker model results in QuickSight D) Analyze A/B test results

Answer: B

Explanation: QuickSight ML Insights provides code-free Anomaly Detection, Forecasting, and Auto-narratives for time-series data. It internally uses Amazon Forecast and the Random Cut Forest algorithm. Business users can obtain ML-driven insights without data scientists.

A) Provides data lake security policy templates B) Automates data ingestion workflows from common sources (RDS, DMS, etc.) to the S3 data lake C) Automatically generates Glue ETL job code D) Provides QuickSight dashboard templates

Answer: B

Explanation: Lake Formation Blueprints provide pre-configured workflows for common ingestion patterns (Database Snapshot, Incremental Database, Log File). They automatically create Glue crawlers and ETL jobs internally, reducing the complexity of data lake construction.

A) Switch the Glue job to Python Shell B) Use the groupFiles option to group small files and increase worker count C) Load data into Redshift first, then unload as Parquet D) Parallelize conversion with a Lambda function chain

Answer: B

Explanation: Glue ETL's groupFiles setting logically groups small files to optimize processing partition count. Increasing the number of workers (DPUs) enables greater parallelism. Spark internally processes data in parallel partitions, so sufficient DPU allocation is needed to fully utilize the parallelism.

A) Check CPUUtilization in CloudWatch metrics B) Analyze STL_QUERY, STL_EXPLAIN, SVL_QUERY_SUMMARY system tables C) Search by query ID in Redshift audit logs D) Trace API calls with CloudTrail

Answer: B

Explanation: Redshift system tables enable query performance analysis. STL_QUERY contains completed query information, STL_EXPLAIN contains query plans, and SVL_QUERY_SUMMARY provides per-step execution statistics. Use the EXPLAIN command to view the execution plan and analyze data distribution, join strategies, and sort key effectiveness.

A) Both conditions must be met before data is delivered to S3 B) Data is delivered to S3 when either condition is met first C) Only Buffer Size is used as the delivery criterion D) Only Buffer Interval is used as the delivery criterion

Answer: B

Explanation: Kinesis Firehose delivers data to S3 when either the Buffer Size (MB) or Buffer Interval (seconds) threshold is reached first. For example, with a 5 MB buffer size and 60-second interval, data is flushed when either 5 MB accumulates or 60 seconds elapse — whichever comes first.

A) All Read operations on the table B) Write change events (Put, Update, Delete) retained for 24 hours C) Only Scan operations D) GSI/LSI creation and deletion events

Answer: B

Explanation: DynamoDB Streams captures item-level changes (INSERT, MODIFY, REMOVE) on the table, retaining them for up to 24 hours. It can capture the before/after images of changed items. Combined with Lambda triggers, it enables event-driven architectures, real-time aggregations, and cross-system synchronization.

A) Move all data to Redshift and analyze there B) AWS Glue Catalog + Athena Federated Query + Lake Formation integrated governance C) Use separate analytics tools for each source D) Consolidate all data into DynamoDB

Answer: B

Explanation: Using the AWS Glue Catalog as a central metadata store, Athena Federated Query for single-SQL access across heterogeneous sources, and Lake Formation for unified security and governance is the most comprehensive solution. Data remains in place; a unified access control policy is applied across all sources.