Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 33
Question 320
A service experiences a regional burst beyond one AZ's capacity. How should capacity be distributed?
- Place stateless application instances in an Auto Scaling group behind an Application Load Balancer.
- Place Auto Scaling instances across multiple suitable Availability Zones behind an ALB.
- Use scheduled Auto Scaling capacity adjustments before predictable peaks.
- Route requests through an ALB DNS endpoint with registered Auto Scaling targets.
Explanation
Correct Answer
B:Multi-AZ capacity spreads demand and reduces dependence on one AZ for new launches. Disable an AZ and inspect remaining target capacity and request success.
Incorrect Answers
A:The ALB distributes requests while the group maintains and scales healthy instances. Place Auto Scaling instances across multiple suitable Availability Zones behind an ALB. Disable an AZ and inspect remaining target capacity and request success.
C:Scheduled scaling prepositions capacity for known demand while dynamic policies handle unexpected fluctuations. Place Auto Scaling instances across multiple suitable Availability Zones behind an ALB. Disable an AZ and inspect remaining target capacity and request success.
D:A load balancer abstracts instance replacement; direct instance IPs defeat fleet elasticity. Multi-AZ capacity spreads demand and reduces dependence on one AZ for new launches.
Question 321
Several stateless API instances repeatedly compute the same frequently requested read-only response. What improves reuse without binding clients to one server?
- Cache eligible responses in a shared ElastiCache store with a defined expiration policy.
- Use deliberate TTL and update-triggered invalidation for affected cache keys.
- Store cart state in DynamoDB or another durable shared datastore.
- Write a durable work event with a transactional outbox and deliver asynchronously.
Explanation
Correct Answer
A:A distributed cache shares results across application instances while TTL bounds staleness. Request the same key through two instances and compare backend reads.
Incorrect Answers
B:Cache lifetimes must match data freshness requirements; an unbounded cache causes stale responses. Cache eligible responses in a shared ElastiCache store with a defined expiration policy. Request the same key through two instances and compare backend reads.
C:Externalized state allows requests to move among disposable workers without losing data. Cache eligible responses in a shared ElastiCache store with a defined expiration policy. Request the same key through two instances and compare backend reads.
D:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Cache eligible responses in a shared ElastiCache store with a defined expiration policy. Request the same key through two instances and compare backend reads.
Question 322
A background email sender must not make a successful database transaction wait for SMTP delivery. Which integration pattern fits?
- Store cart state in DynamoDB or another durable shared datastore.
- Publish an asynchronous event to downstream analytics consumers outside the request path.
- Record an outbox event atomically with the business transaction and publish it asynchronously.
- Write a durable work event with a transactional outbox and deliver asynchronously.
Explanation
Correct Answer
D:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Simulate a process crash after commit and verify the email event is eventually dispatched once logically.
Incorrect Answers
A:Externalized state allows requests to move among disposable workers without losing data. Write a durable work event with a transactional outbox and deliver asynchronously. Simulate a process crash after commit and verify the email event is eventually dispatched once logically.
B:Separating optional side effects from critical request latency reduces coupling during analytics outages. Write a durable work event with a transactional outbox and deliver asynchronously. Simulate a process crash after commit and verify the email event is eventually dispatched once logically.
C:Transactionally storing both business change and pending event prevents lost updates on intermediate process crashes. Write a durable work event with a transactional outbox and deliver asynchronously. Simulate a process crash after commit and verify the email event is eventually dispatched once logically.
Question 323
An application inserts a record in its database and then publishes an event; a crash between steps silently loses notifications. Which design repairs the dual-write gap?
- Record an outbox event atomically with the business transaction and publish it asynchronously.
- Cache eligible responses in a shared ElastiCache store with a defined expiration policy.
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Publish an asynchronous event to downstream analytics consumers outside the request path.
Explanation
Correct Answer
A:Transactionally storing both business change and pending event prevents lost updates on intermediate process crashes. Crash after commit and verify the event can still be published from the outbox.
Incorrect Answers
B:A distributed cache shares results across application instances while TTL bounds staleness. Record an outbox event atomically with the business transaction and publish it asynchronously. Crash after commit and verify the event can still be published from the outbox.
C:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Record an outbox event atomically with the business transaction and publish it asynchronously. Crash after commit and verify the event can still be published from the outbox.
D:Separating optional side effects from critical request latency reduces coupling during analytics outages. Record an outbox event atomically with the business transaction and publish it asynchronously. Crash after commit and verify the event can still be published from the outbox.
Question 324
Multiple app instances hold mutable shopping carts in process memory, and carts vanish on replacement. What change supports instance independence?
- Store cart state in DynamoDB or another durable shared datastore.
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Use deliberate TTL and update-triggered invalidation for affected cache keys.
- Record an outbox event atomically with the business transaction and publish it asynchronously.
Explanation
Correct Answer
A:Externalized state allows requests to move among disposable workers without losing data. Replace all web instances and retrieve an existing cart using its customer ID.
Incorrect Answers
B:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Store cart state in DynamoDB or another durable shared datastore. Replace all web instances and retrieve an existing cart using its customer ID.
C:Cache lifetimes must match data freshness requirements; an unbounded cache causes stale responses. Store cart state in DynamoDB or another durable shared datastore. Replace all web instances and retrieve an existing cart using its customer ID.
D:Transactionally storing both business change and pending event prevents lost updates on intermediate process crashes. Externalized state allows requests to move among disposable workers without losing data.
Question 325
A cache serves customer catalog values for hours after price updates. What cache coherency mechanism is missing?
- Use deliberate TTL and update-triggered invalidation for affected cache keys.
- Cache eligible responses in a shared ElastiCache store with a defined expiration policy.
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Record an outbox event atomically with the business transaction and publish it asynchronously.
Explanation
Correct Answer
A:Cache lifetimes must match data freshness requirements; an unbounded cache causes stale responses. Update a product price and confirm cache replacement or expiry within the service objective.
Incorrect Answers
B:A distributed cache shares results across application instances while TTL bounds staleness. Use deliberate TTL and update-triggered invalidation for affected cache keys. Update a product price and confirm cache replacement or expiry within the service objective.
C:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Use deliberate TTL and update-triggered invalidation for affected cache keys. Update a product price and confirm cache replacement or expiry within the service objective.
D:Transactionally storing both business change and pending event prevents lost updates on intermediate process crashes. Cache lifetimes must match data freshness requirements; an unbounded cache causes stale responses.
Question 326
Thousands of containers use the same database with too many independent connections. Which middleware reduces connection pressure?
- Use an appropriate managed connection pool or RDS Proxy for supported databases.
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Cache eligible responses in a shared ElastiCache store with a defined expiration policy.
- Use stable service discovery names or managed routing endpoints instead of instance addresses.
Explanation
Correct Answer
A:A proxy pools reusable connections so fleet growth does not linearly grow database connection counts. Scale containers and inspect peak database connections and proxy borrowing.
Incorrect Answers
B:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Use an appropriate managed connection pool or RDS Proxy for supported databases. Scale containers and inspect peak database connections and proxy borrowing.
C:A distributed cache shares results across application instances while TTL bounds staleness. Use an appropriate managed connection pool or RDS Proxy for supported databases. Scale containers and inspect peak database connections and proxy borrowing.
D:Logical service endpoints allow backends to scale and roll independently of callers. Use an appropriate managed connection pool or RDS Proxy for supported databases. Scale containers and inspect peak database connections and proxy borrowing.
Question 327
A critical request needs a response immediately, but optional analytics processing takes much longer. Where should analytics execute?
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Cache eligible responses in a shared ElastiCache store with a defined expiration policy.
- Publish an asynchronous event to downstream analytics consumers outside the request path.
- Store cart state in DynamoDB or another durable shared datastore.
Explanation
Correct Answer
C:Separating optional side effects from critical request latency reduces coupling during analytics outages. Pause analytics and verify critical requests still complete while events remain available.
Incorrect Answers
A:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Publish an asynchronous event to downstream analytics consumers outside the request path. Pause analytics and verify critical requests still complete while events remain available.
B:A distributed cache shares results across application instances while TTL bounds staleness. Publish an asynchronous event to downstream analytics consumers outside the request path. Pause analytics and verify critical requests still complete while events remain available.
D:Externalized state allows requests to move among disposable workers without losing data. Separating optional side effects from critical request latency reduces coupling during analytics outages.
Question 328
Several services reference hardcoded hostnames that change with every deployment. What abstraction reduces deployment coupling?
- Use an appropriate managed connection pool or RDS Proxy for supported databases.
- Write a durable work event with a transactional outbox and deliver asynchronously.
- Use stable service discovery names or managed routing endpoints instead of instance addresses.
- Record an outbox event atomically with the business transaction and publish it asynchronously.
Explanation
Correct Answer
C:Logical service endpoints allow backends to scale and roll independently of callers. Replace service targets without changing client configuration.
Incorrect Answers
A:A proxy pools reusable connections so fleet growth does not linearly grow database connection counts. Use stable service discovery names or managed routing endpoints instead of instance addresses. Replace service targets without changing client configuration.
B:An outbox avoids the dual-write gap between committing business state and queuing downstream work. Use stable service discovery names or managed routing endpoints instead of instance addresses. Replace service targets without changing client configuration.
D:Transactionally storing both business change and pending event prevents lost updates on intermediate process crashes. Logical service endpoints allow backends to scale and roll independently of callers.
Question 329
Sensor events arrive continuously at high volume and must be consumed independently by fraud and analytics systems. Which streaming backbone fits?
- Use Kinesis enhanced fan-out for the dedicated registered consumer.
- Use Kinesis Data Streams with separate consumer applications.
- Persist records in a retained Kinesis Data Stream and process them when consumers return.
- Increase Kinesis stream retention to cover the required replay window.
Explanation
Correct Answer
B:A retained append-only stream lets multiple consumers process the event sequence at independent rates. Pause fraud processing and verify analytics remains current and fraud resumes within retention.
Incorrect Answers
A:Enhanced fan-out provides consumer-specific throughput paths rather than sharing ordinary shard read capacity. Use Kinesis Data Streams with separate consumer applications. Pause fraud processing and verify analytics remains current and fraud resumes within retention.
C:A stream buffers records independently of consumer availability within retention limits. Use Kinesis Data Streams with separate consumer applications. Pause fraud processing and verify analytics remains current and fraud resumes within retention.
D:Kinesis retains records for a configured interval so recovery depends on that window. Use Kinesis Data Streams with separate consumer applications. Pause fraud processing and verify analytics remains current and fraud resumes within retention.
Question 330
One tenant's hot partition key dominates a Kinesis stream shard and throttles writes. What should be revised?
- Distribute partition keys more evenly across available shards while preserving required ordering groups.
- Persist records in a retained Kinesis Data Stream and process them when consumers return.
- Choose the customer ID as an appropriate Kinesis partition key with sufficient shard distribution.
- Increase Kinesis stream retention to cover the required replay window.
Explanation
Correct Answer
A:Partition-key hashing determines shard load; a hot key cannot be fixed by adding consumers alone. Measure per-shard incoming records and compare after repartitioning.
Incorrect Answers
B:A stream buffers records independently of consumer availability within retention limits. Distribute partition keys more evenly across available shards while preserving required ordering groups. Measure per-shard incoming records and compare after repartitioning.
C:Records sharing a partition key reach the same shard and preserve per-shard order; overall global order is not guaranteed. Distribute partition keys more evenly across available shards while preserving required ordering groups. Measure per-shard incoming records and compare after repartitioning.
D:Kinesis retains records for a configured interval so recovery depends on that window. Distribute partition keys more evenly across available shards while preserving required ordering groups. Measure per-shard incoming records and compare after repartitioning.