Question 90
A small data API uses Lambda and an API Gateway route, and developers want concise repeatable deployments. What framework fits?
- AWS CDK construct
- AWS CloudFormation stack
- CloudFormation change set
- AWS SAM template
Explanation
Correct Answer
D:AWS SAM extends CloudFormation with serverless resource declarations for applications; in this scenario, declare the serverless resources in a SAM template, build dependencies and deploy through the managed CloudFormation stack.
Incorrect Answers
A:AWS CDK expresses cloud infrastructure using supported programming languages that synthesize CloudFormation; this helps with reusing programmable infrastructure definitions across teams, not packaging Lambda and Step Functions assets as serverless infrastructure.
B:CloudFormation provisions and updates AWS resources through declarative templates and stack change tracking; although useful for recreating the same data pipeline infrastructure in different accounts, it leaves the need for packaging Lambda and Step Functions assets as serverless infrastructure unmet.
C:A change set previews proposed modifications to a stack before they are executed; it would target reviewing potentially destructive infrastructure updates, while the question is about packaging Lambda and Step Functions assets as serverless infrastructure.
Question 91
A Glue job only needs to list an incoming bucket and read keys under `incoming/orders/`. What permission design is least privilege?
- IAM permissions boundary
- IAM explicit deny precedence
- IAM resource-scoped S3 policy
- IAM policy condition key
Explanation
Correct Answer
C:An IAM identity policy can scope supported S3 operations to specific bucket and object ARN patterns; in this scenario, separate bucket-level listing conditions from object-level GetObject ARNs so the role cannot read unrelated prefixes.
Incorrect Answers
A:A permissions boundary limits the maximum effective permissions of a principal even when identity policies allow more; its benefit is constraining delegated ETL-role administration; that is different from granting ETL read access to one landing prefix.
B:An explicit deny in an applicable policy overrides otherwise allowing IAM statements; the current need is for granting ETL read access to one landing prefix, not diagnosing access denied despite an Allow statement.
D:IAM policy conditions can limit allowed API use based on supported request attributes and context; this helps with restricting object writes to approved encryption settings, not granting ETL read access to one landing prefix.
Question 92
An event feed adds an optional promotion code, and historical Parquet files have no such field. Which evolution strategy is least disruptive?
- Schema-on-read normalization
- Schema registry compatibility check
- Controlled schema migration version
- Backward-compatible optional column
Explanation
Correct Answer
D:Adding an optional column can allow older records and consumers to coexist with newer schema versions; in this scenario, add the field with compatible nullable semantics, then verify how each query engine reads absent historic values.
Incorrect Answers
A:Schema-on-read applies structure when data is queried or transformed rather than forcing one physical shape at ingestion; this can support retaining unpredictable third-party fields for later curation, but would not address the need for introducing a nullable attribute without breaking old data.
B:A schema registry can enforce defined compatibility rules for producers and consumers sharing encoded records; this helps with blocking incompatible event schema changes during deployment, not introducing a nullable attribute without breaking old data.
C:Versioned schema migration scripts make database changes reproducible and reviewable across environments; although useful for coordinating relational schema changes with ETL releases, it leaves the need for introducing a nullable attribute without breaking old data unmet.
Question 93
A nightly job needs to move hundreds of gigabytes of compressed files from S3 into a Redshift staging table. Which SQL command fits?
- Redshift UNLOAD to S3
- Redshift COPY manifest
- Redshift staging table merge
- Redshift COPY from S3
Explanation
Correct Answer
D:COPY loads supported data files from S3 into Redshift tables using managed parallel reads; in this scenario, use COPY with an authorized IAM role and suitable input format, distributing load files for efficient parallel ingestion.
Incorrect Answers
A:UNLOAD writes Redshift query results to S3 objects for exchange or downstream analysis; this helps with exporting a curated warehouse query result to the lake, not ingesting a large staged dataset into warehouse tables.
B:A COPY manifest specifies the exact objects that a load should consume for reproducibility; although useful for preventing partial object sets during a warehouse load, it leaves the need for ingesting a large staged dataset into warehouse tables unmet.
C:Loading into a staging table before a controlled merge separates ingestion from publication of target data; it would target atomically applying validated changes to warehouse tables, while the question is about ingesting a large staged dataset into warehouse tables.
Question 94
An upstream API permits only twenty concurrent requests, but the collector starts 200 threads. Which setting addresses the constraint?
- Client-side concurrency limit
- Kinesis on-demand capacity
- RDS Proxy connection pooling
- Exponential backoff with jitter
- DynamoDB adaptive capacity
Explanation
Correct Answer
A:A semaphore or worker cap bounds simultaneous API requests sent by an ingestion job; in this scenario, limit in-flight calls independently of retry delays, keeping concurrency within the contractual service capacity.
Incorrect Answers
B:Kinesis on-demand mode adjusts stream capacity without manually planning shard counts for typical demand patterns; although useful for absorbing variable streaming ingestion demand without shard planning, it leaves the need for keeping API requests below a known in-flight limit unmet.
C:RDS Proxy pools application database connections and reduces connection churn against supported databases; it would target protecting a relational database from short-lived connection bursts, while the question is about keeping API requests below a known in-flight limit.
D:Backoff and jitter distribute API retries in time after throttling responses or transient errors; its benefit is preventing synchronized retries after a shared API quota is exhausted; that is different from keeping API requests below a known in-flight limit.
E:DynamoDB adaptive capacity can redistribute available throughput to heavily accessed partitions within service limits; the current need is for keeping API requests below a known in-flight limit, not mitigating uneven partition access within a DynamoDB table.
Question 95
An operations team has a consistent incident-response function and new members need the same minimal permissions. Which authorization model fits?
- Cross-account bucket policy
- Lake Formation LF-Tag based access
- IAM ABAC with principal tags
- Session tag authorization
- S3 access point policy
- IAM RBAC role assignment
Explanation
Correct Answer
F:Role-based access control groups permissions according to job functions or workload responsibilities; in this scenario, attach reviewed policies to an operator role and control who can assume it, avoiding per-person ad hoc grants.
Incorrect Answers
A:An S3 bucket resource policy can allow a designated external account or role to access selected objects; this can support authorizing direct read access from another AWS account, but would not address the need for giving production operators a stable predefined permission set.
B:LF-Tag authorization grants governed table access based on metadata classifications and expressions; this helps with authorizing lake assets by governed data classification, not giving production operators a stable predefined permission set.
C:Attribute-based access control compares principal and resource attributes through supported IAM condition keys; although useful for mapping ETL roles to same-team tagged resources, it leaves the need for giving production operators a stable predefined permission set unmet.
D:Session tags can carry approved attributes in assumed-role sessions for ABAC decisions; it would target propagating per-session team attributes to authorization rules, while the question is about giving production operators a stable predefined permission set.
E:S3 access points offer distinct access controls and network policies for shared S3 datasets; its benefit is exposing one bucket to applications with different data-access boundaries; that is different from giving production operators a stable predefined permission set.
Question 96
A large dataset repeats only five status strings across millions of rows. What Parquet encoding technique can reduce its footprint?
- Columnar encoding dictionary
- Avoiding overpartitioned Spark output
- Sorting output by query predicates
- S3 object compaction
Explanation
Correct Answer
A:Dictionary encoding can efficiently represent repeated low-cardinality values in columnar storage formats; in this scenario, dictionary encoding represents repeated values by compact codes where the format and engine choose it effectively.
Incorrect Answers
B:Excessive Spark partitions can create scheduling overhead and tiny output objects on low-volume workloads; it would target tuning parallel output granularity to data volume, while the question is about compressing repeated categorical attributes efficiently.
C:Sorting records on frequently filtered fields can improve locality and block skipping for some engines; its benefit is increasing pruning efficiency inside large files; that is different from compressing repeated categorical attributes efficiently.
D:Compaction combines many small analytic files into fewer appropriately sized objects; the current need is for compressing repeated categorical attributes efficiently, not reducing excessive query task overhead.
Question 97
A Glue run fails immediately with a generic console status but no obvious data corruption. What is the first troubleshooting source?
- Spark shuffle skew investigation
- Data type casting validation
- Executor memory tuning
- Glue error log inspection
Explanation
Correct Answer
D:Glue job failures produce execution logs that distinguish syntax, credentials, schema and runtime failures; in this scenario, open the job's CloudWatch logs and job run details to find the exception and failing stage before changing worker settings.
Incorrect Answers
A:Spark stage metrics can reveal straggler tasks and imbalanced shuffle partitions in a join; this helps with finding a slow join caused by one hot key, not locating the precise failing operation in Glue.
B:Explicit type validation and casting prevent malformed input strings from corrupting typed transformations; although useful for handling unexpected nonnumeric values in a numeric column, it leaves the need for locating the precise failing operation in Glue unmet.
C:Spark executor memory and partition size affect whether large tasks spill or fail from memory pressure; it would target addressing executor out-of-memory errors on large partitions, while the question is about locating the precise failing operation in Glue.
Question 98
A new supplier feed has unknown missing-value and outlier patterns. Which DataBrew operation should run before writing transformations?
- DataBrew profiling job
- Glue DataBrew recipe
- Lambda event-driven preprocessing
- SageMaker Unified Studio data exploration
- Glue Spark preprocessing job
Explanation
Correct Answer
A:A DataBrew profile job measures dataset properties including value patterns, completeness and distributions; in this scenario, profile the dataset first to obtain evidence about field distributions and invalid data, then design cleaning rules.
Incorrect Answers
B:DataBrew recipes apply documented preparation steps such as normalizing and cleaning tabular fields; the current need is for inspecting quality problems before deciding cleaning rules, not letting analysts prepare messy CSV columns without writing Spark code.
C:Lambda can perform small validation or normalization tasks when an object-arrival event is received; this can support normalizing small files immediately after landing, but would not address the need for inspecting quality problems before deciding cleaning rules.
D:Unified Studio provides a governed collaborative environment for exploring and preparing datasets; this helps with preparing datasets in a shared analytics project, not inspecting quality problems before deciding cleaning rules.
E:Glue Spark jobs transform larger datasets with joins, repartitioning and schema mapping; although useful for preparing large multi-file datasets for analytics, it leaves the need for inspecting quality problems before deciding cleaning rules unmet.
Question 99
Several Lambda functions need access to the same shared reference files, exceeding feasible per-invocation downloads. Which storage integration fits?
- Distributed partitioned compute
- Lambda /tmp ephemeral storage
- Lambda Amazon EFS mount
- Graph traversal algorithm
Explanation
Correct Answer
C:Lambda can mount an EFS file system through access points from appropriately networked functions; in this scenario, configure an EFS access point and VPC connectivity so functions use persistent shared files within their permissions.
Incorrect Answers
A:Distributed processing partitions data so workers handle portions in parallel and combine results as needed; the current need is for sharing persistent files among multiple Lambda executions, not parallelizing a large aggregation beyond one machine.
B:Lambda provides configurable temporary /tmp storage local to an execution environment; this can support buffering intermediate files during one function execution, but would not address the need for sharing persistent files among multiple Lambda executions.
D:Graph traversal explores connected vertices and edges to follow relationships or dependencies; although useful for following lineage dependencies between datasets, it leaves the need for sharing persistent files among multiple Lambda executions unmet.
Question 100
A retailer needs to land incoming purchase events in S3 with managed buffering and minimal stream-consumer code. Which service should receive the events?
- Amazon MSK
- DynamoDB Streams
- Amazon Kinesis Data Streams
- Amazon Data Firehose
Explanation
Correct Answer
D:Firehose buffers streaming records and delivers them to configured analytics destinations; in this scenario, configure an S3 destination and appropriate buffering; delivery is managed without maintaining custom shard readers.
Incorrect Answers
A:Amazon MSK runs managed Apache Kafka brokers while retaining Kafka client and topic semantics; the current need is for continuously depositing events into an S3 data lake, not migrating existing Kafka producers without rewriting clients.
B:DynamoDB Streams exposes item-level changes made to a DynamoDB table for downstream processors; this can support reacting to inserts and updates of table items, but would not address the need for continuously depositing events into an S3 data lake.
C:A data stream retains ordered records in shards for applications that read at their own pace; this helps with keeping a replayable clickstream for independent consumers, not continuously depositing events into an S3 data lake.