Question 60
A vendor allows only approved public source IPs; Glue connections originate from private subnets and access its public API. Which egress design helps?
- Security group inbound rule
- S3 gateway endpoint
- NAT gateway stable egress IP
- VPC interface endpoint
Explanation
Correct Answer
C:A NAT gateway in a public subnet can provide predictable public egress through its Elastic IP; in this scenario, route outbound traffic through the NAT gateway and register its Elastic IP with the vendor; retain private IP space for workers.
Incorrect Answers
A:A security group rule allows specified inbound protocols, ports and sources to a resource interface; although useful for allowing a Glue job to reach a private JDBC database, it leaves the need for providing fixed outbound addresses for a private-subnet API poller unmet.
B:An S3 gateway endpoint adds a VPC route to S3 through the AWS private network path; it would target reading S3 objects from a private subnet without a NAT gateway, while the question is about providing fixed outbound addresses for a private-subnet API poller.
D:An interface VPC endpoint offers private connectivity to supported AWS service APIs using endpoint ENIs; the current need is for providing fixed outbound addresses for a private-subnet API poller, not avoiding public internet transit for a supported AWS API.
Question 61
Producers report intermittent Kinesis write failures even when average stream throughput looks low. Which evidence should be checked?
- Athena scanned-data metrics
- Glue Spark stage metrics
- Redshift EXPLAIN plan
- Kinesis shard throughput metrics
Explanation
Correct Answer
D:Stream shard metrics reveal write or read throughput saturation and throttling within shard limits; in this scenario, inspect shard-level throughput and throttling; concentrated partition-key traffic can exhaust one shard before aggregate capacity.
Incorrect Answers
A:Athena query execution statistics report the volume of data scanned and runtime behavior; this can support explaining a sudden rise in query cost, but would not address the need for diagnosing write throttling on a hot partition key.
B:Spark task and stage metrics reveal shuffle size, skew, spill and slow executors; this helps with identifying why one Spark stage dominates the run, not diagnosing write throttling on a hot partition key.
C:EXPLAIN shows a planned Redshift query strategy, including joins, scans and potential data movement; although useful for identifying costly redistribution in warehouse SQL, it leaves the need for diagnosing write throttling on a hot partition key unmet.
Question 62
Mobile devices send short JSON events over HTTPS and require authentication plus request throttling before ingestion. Which front end should be chosen?
- Secrets Manager for API tokens
- Checkpointed API pagination
- API Gateway with Lambda
- Lambda scheduled API poller
Explanation
Correct Answer
C:API Gateway accepts HTTPS requests and can invoke Lambda for backend request processing; in this scenario, use an API with an authorizer and throttling policy before forwarding accepted payloads to downstream processing.
Incorrect Answers
A:Secrets Manager stores API credentials and can support rotation workflows for supported secrets; this can support managing third-party API credentials without embedding them in code, but would not address the need for exposing a managed HTTPS entry point for ingested events.
B:A durable cursor or page token enables an ingestion process to resume an API listing incrementally; this helps with resuming paginated ingestion after partial failures, not exposing a managed HTTPS entry point for ingested events.
D:Lambda code can invoke third-party REST APIs on a schedule when network access is available; it would target polling a vendor API at bounded intervals, while the question is about exposing a managed HTTPS entry point for ingested events.
Question 63
A PostgreSQL-backed catalog must rank nearby document embeddings while retaining SQL joins to metadata. What index approach fits?
- Aurora PostgreSQL pgvector HNSW index
- Amazon ElastiCache cache
- Amazon OpenSearch Service search index
- Amazon MemoryDB for Redis
Explanation
Correct Answer
A:Aurora PostgreSQL with supported pgvector functionality can index embeddings for approximate similarity search; in this scenario, use a supported vector index such as HNSW on the embedding column and tune distance operator and index parameters.
Incorrect Answers
B:ElastiCache provides managed in-memory caching to accelerate access to data stored durably elsewhere; it would target caching read-heavy reference values without changing the source of truth, while the question is about finding similar embeddings within relational application data.
C:OpenSearch Service supports full-text search and search-oriented indexing of documents; its benefit is enabling relevance-ranked keyword search over ingested articles; that is different from finding similar embeddings within relational application data.
D:MemoryDB offers a durable Redis-compatible in-memory database for low-latency key-value access; the current need is for finding similar embeddings within relational application data, not combining Redis-compatible operations with a durable primary database.
Question 64
A nightly ETL job must read a MySQL source in private subnets and convert records before landing them in S3. Which ingestion connector is suitable?
- Amazon AppFlow
- AWS DMS full load
- S3 object batch processing with Glue
- AWS Glue JDBC connection
Explanation
Correct Answer
D:Glue can read JDBC-backed sources through a connection with network and credential settings; in this scenario, configure the JDBC URL, credentials and VPC connectivity for Glue so its job can retrieve the relational source rows.
Incorrect Answers
A:AppFlow transfers data between supported SaaS applications and AWS destinations with managed connectors; this can support importing CRM objects from a supported SaaS application, but would not address the need for extracting scheduled rows from a private relational database.
B:AWS DMS can copy existing database tables as an initial migration or replication phase; this helps with taking a one-time baseline of source tables, not extracting scheduled rows from a private relational database.
C:Glue batch ETL processes stored S3 objects when a scheduled job runs against their paths; although useful for reprocessing a monthly collection of stored flat files, it leaves the need for extracting scheduled rows from a private relational database unmet.
Question 65
Raw CSV logs in S3 are repeatedly queried for only two of fifty columns, driving scan charges. Which storage format should ETL produce?
- Compression-aware partitioned Parquet
- JSON preservation for variable schema
- Apache Parquet output
- Apache Avro for row-oriented exchange
Explanation
Correct Answer
C:Parquet stores column-oriented data with encoding and compression suitable for selective analytics; in this scenario, convert to Parquet with appropriate compression and schema so query engines can skip nonreferenced column chunks.
Incorrect Answers
A:Compression and sensible file sizing improve Parquet read efficiency without creating countless tiny files; its benefit is avoiding a data lake filled with tiny output objects; that is different from reducing scan bytes for repeated analytic queries.
B:JSON can preserve nested and changing attributes that are awkward in a rigid flat-row schema; the current need is for reducing scan bytes for repeated analytic queries, not preserving semi-structured event attributes during initial landing.
D:Avro serializes row-oriented records with schema-based encoding suitable for data interchange; this helps with exchanging schema-described records across heterogeneous consumers, not reducing scan bytes for repeated analytic queries.
Question 66
A private RDS database should accept JDBC traffic from Glue workers but not from other VPC workloads. What control should change?
- Private VPC endpoint policy
- Database security-group ingress
- Glue worker security-group egress
- Network ACL rule order
Explanation
Correct Answer
B:A VPC security group can restrict database listener access by source security group and destination port; in this scenario, allow only the Glue worker security-group source on the database port and verify routing, DNS and NACL behavior.
Incorrect Answers
A:VPC endpoint policies can restrict supported API actions and resource access through a particular endpoint; this can support limiting S3 API access through a specific gateway endpoint, but would not address the need for permitting only Glue ETL workers to reach RDS over TCP.
C:Security-group egress rules govern which destinations Glue worker ENIs may contact; although useful for allowing a job to initiate approved outbound database connections, it leaves the need for permitting only Glue ETL workers to reach RDS over TCP unmet.
D:Network ACLs are stateless subnet controls with numbered allow and deny rules evaluated in order; it would target debugging return traffic dropped at a subnet boundary, while the question is about permitting only Glue ETL workers to reach RDS over TCP.
Question 67
An RDS ingestion transaction holds locks during a lengthy network call, blocking online customer updates. What design change helps?
- Shorter database transaction
- Redshift lock monitoring
- Relational isolation-level selection
- Partitioned load scheduling
Explanation
Correct Answer
A:Reducing transaction length releases row and table locks sooner in conventional relational engines; in this scenario, commit coherent batches promptly and perform remote calls outside critical transactions where consistency permits.
Incorrect Answers
B:Redshift system views and query monitoring expose sessions waiting on resource or transaction locks; its benefit is diagnosing warehouse query waits caused by locks; that is different from limiting lock contention from long-running batch updates.
C:Transaction isolation controls the concurrency anomalies a database transaction permits under its engine semantics; the current need is for limiting lock contention from long-running batch updates, not choosing consistency guarantees for simultaneous source updates.
D:Loading separate table partitions at coordinated intervals can reduce contention with user-facing operations; this can support avoiding bulk-load conflicts with live warehouse queries, but would not address the need for limiting lock contention from long-running batch updates.
Question 68
A team runs exploratory SQL against S3 lake objects only twice per month and wants to minimize standing warehouse infrastructure. Which query model is best suited?
- Lambda short execution
- Redshift Serverless
- Athena on-demand SQL
- Provisioned Redshift cluster
- EMR persistent cluster
Explanation
Correct Answer
C:Athena can query data in S3 without administering a persistent warehouse cluster; in this scenario, use an appropriate Athena capacity or on-demand arrangement and manage query efficiency through columnar layout and pruning.
Incorrect Answers
A:Lambda charges for invoked execution and suits bounded tasks under platform runtime constraints; it would target handling sporadic lightweight event transformations without idle servers, while the question is about serving occasional lake SQL queries with variable demand.
B:Redshift Serverless runs warehouse workloads without manually managing provisioned clusters; its benefit is running variable warehouse SQL workloads with less infrastructure management; that is different from serving occasional lake SQL queries with variable demand.
D:Provisioned Redshift compute provides explicit sizing and reservations for more predictable warehouse workloads; this can support optimizing sustained analytics demand under stable capacity, but would not address the need for serving occasional lake SQL queries with variable demand.
E:A persistent EMR cluster can run continuing workloads with custom node and framework configurations; this helps with maintaining specialized continuously running Spark infrastructure, not serving occasional lake SQL queries with variable demand.
Question 69
A Redshift query repeatedly shuffles a huge fact table to join on tenant_id with another large table. What physical design merits review?
- S3 partition cardinality
- Redshift distribution key choice
- Relational composite index order
- Parquet compression codec
Explanation
Correct Answer
B:Redshift distribution style influences where rows reside and how much data moves during joins; in this scenario, co-locate commonly joined data where suitable using distribution strategy, balancing skew and changes in query workload.
Incorrect Answers
A:Partitioning by extremely high-cardinality fields can create excessive small partitions and metadata overhead; its benefit is avoiding millions of tiny lake partitions; that is different from reducing redistribution for frequent large-table joins.
C:The order of columns in a composite B-tree index affects which query predicates can use its leading keys; this can support optimizing multi-column predicates in an operational database, but would not address the need for reducing redistribution for frequent large-table joins.
D:A compression codec changes encoded data size and decompression trade-offs for columnar lake files; this helps with reducing bytes stored and scanned for Parquet files, not reducing redistribution for frequent large-table joins.
Question 70
Marketing IDs and payment IDs use different numbering schemes, producing mismatched records in a join. What design step is needed?
- Deduplication across sources
- Late-arriving dimension handling
- Time zone normalization
- Stable join key
Explanation
Correct Answer
D:Joining data from separate systems requires a trustworthy key that identifies the same entity on each side; in this scenario, create or use an authoritative crosswalk key, validate uniqueness and normalize source identifiers before joining.
Incorrect Answers
A:Entity deduplication uses an agreed business key and conflict-resolution rule across integrated sources; this can support merging duplicate customer updates from two systems, but would not address the need for preventing incorrect joins across customer feeds.
B:Late-arriving dimensions require a defined strategy for facts received before their lookup rows; this helps with loading facts that precede their dimension records, not preventing incorrect joins across customer feeds.
C:Normalizing timestamps to a common time basis prevents false ordering across systems using different zones; although useful for correlating event timestamps from multiple geographic sources, it leaves the need for preventing incorrect joins across customer feeds unmet.