Design for New Solutions practice questions

Amazon ElastiCache for Redis

Why wrong: Amazon ElastiCache for Redis provides sub-millisecond latency but is not serverless; it requires managing cache clusters, which conflicts with the serverless requirement. DynamoDB, with its serverless nature and TTL support, is a better fit.

Amazon RDS for MySQL

Why wrong: Amazon RDS for MySQL is a relational database with higher latency and is not serverless, making it unsuitable for session state in a serverless application.

Amazon S3

Why wrong: Amazon S3 is an object storage service with high latency for frequent read/write operations, not designed for session state.

The S3 bucket policy denies access to the IAM role.

Why wrong: No bucket policy is shown; the issue is in the IAM policy.

The IAM role is not attached to the EC2 instance profile.

Why wrong: The exhibit shows the role has the policy; the issue is likely the condition.

The policy lacks permissions for s3:ListBucket.

Why wrong: The action is s3:GetObject, which is sufficient for downloading objects.

Use cross-region read replicas with synchronous replication.

Why wrong: Aurora does not support synchronous cross-region replication. This option is invalid.

Deploy Aurora Global Database with multi-master clusters in two regions.

Why wrong: Amazon Aurora Global Database does not support multi-master clusters. It uses a single primary with read-only secondaries. This option is incorrect.

Use Amazon DynamoDB Global Tables instead of Aurora.

Why wrong: DynamoDB Global Tables can meet the RPO and RTO requirements, but the question specifically requires using Aurora Global Database. Therefore, this is not a valid configuration within the given constraints.

S3 Intelligent-Tiering

Why wrong: S3 Intelligent-Tiering automatically moves data between frequent and infrequent access tiers, but it charges monitoring fees and can incur retrieval costs if data is moved to the infrequent access tier. These additional costs may increase total expenses, and storage class does not affect Athena query costs. Hence, it is not the best option for minimizing query costs.

S3 Glacier Deep Archive

Why wrong: S3 Glacier Deep Archive is not suitable for frequent querying due to retrieval times ranging from hours to days and high retrieval costs. It does not minimize query costs.

S3 One Zone-IA

Why wrong: S3 One Zone-IA has lower storage costs but incurs retrieval fees and is less durable. It is not optimal for minimizing query costs in a data lake environment.

Configure an RDS Multi-AZ DB cluster with a standby instance in another Region.

Why wrong: Incorrect. RDS Multi-AZ DB clusters do not support a standby instance in another Region for MySQL; cross-Region synchronous replication is not available.

Use EC2 Auto Scaling to launch new instances in another Region and use an Application Load Balancer.

Why wrong: Incorrect. This addresses compute scalability but does not replicate the database, failing to meet the RPO of 5 seconds.

Take automated snapshots of the RDS instance every 5 seconds and copy them to another Region.

Why wrong: Incorrect. Automated snapshots cannot be taken every 5 seconds; the minimum interval is 5 minutes, which would result in an RPO of at least 5 minutes.

Amazon DocumentDB

Why wrong: DocumentDB is MongoDB-compatible, but not the best fit for high availability beyond single region.

Amazon Neptune

Why wrong: Neptune is graph database, not general NoSQL.

Amazon RDS

Why wrong: RDS is relational, not NoSQL.

The Multi-AZ standby was promoted due to a failover, causing a brief inconsistency.

Why wrong: Multi-AZ failover is for high availability and would not cause ongoing stale reads.

The application is not using a connection pool, causing connections to be established to different replicas.

Why wrong: Connection pooling does not affect consistency; the issue is eventual consistency.

The read replica is using a different storage type (Aurora I/O-Optimized) than the primary.

Why wrong: They are using RDS PostgreSQL, not Aurora.

Configure the SQS DLQ with a redrive policy that allows messages to be sent back to the source queue after 3 retries.

Why wrong: The redrive policy is configured on the source queue, not on the DLQ.

Configure the SQS queue with a visibility timeout of 30 seconds and a redrive policy with maxReceiveCount of 3, pointing to a DLQ.

Why wrong: A 30-second visibility timeout is too short for typical Lambda processing; the visibility timeout should be at least 6 minutes to allow the Lambda function time to process and delete or retry the message.

Configure Lambda with a reserved concurrency of 1 and set the SQS queue's redrive policy to maxReceiveCount of 3.

Why wrong: Reserved concurrency is unrelated to retry behavior and would limit the Lambda function's ability to scale.

Amazon S3 with S3 File Gateway to present as a file system.

Why wrong: Amazon S3 is object storage, not a file system. S3 File Gateway provides a file interface but does not offer native NFS across AZs with IAM authorization like EFS.

Amazon EBS with a Provisioned IOPS volume type, mounted on multiple instances using a cluster file system.

Why wrong: Amazon EBS volumes are designed for single-instance attachment in a single Availability Zone and cannot be shared across multiple instances in different AZs without a cluster file system that still has limitations.

Amazon FSx for Lustre with a persistent deployment type.

Why wrong: Amazon FSx for Lustre is optimized for high-performance computing workloads and does not provide general-purpose NFS with IAM authorization.

Use a single Amazon SQS standard queue with a deduplication ID to ensure each message is processed only once.

Why wrong: Incorrect. SQS standard queues do not guarantee order or exactly-once processing, even with deduplication ID.

Use Amazon Kinesis Data Streams with a single shard and implement idempotent processing in the consumer.

Why wrong: Incorrect. While Kinesis Data Streams with a single shard can provide ordering, it does not guarantee exactly-once processing and introduces additional complexity.

Use Amazon SNS with a FIFO topic and subscribe SQS FIFO queues to ensure ordered delivery.

Why wrong: Incorrect. Although SNS FIFO topics do provide ordering and exactly-once delivery to subscribed SQS FIFO queues, this solution adds unnecessary components and cost when SQS FIFO alone suffices.

Migrate the session store from DynamoDB to Amazon ElastiCache Memcached.

Why wrong: Requires significant application changes and Memcached is not durable.

Increase the read capacity units (RCU) of the DynamoDB table to handle peak traffic.

Why wrong: Manual scaling is not cost-effective and may still lead to throttling if spikes exceed provisioned capacity.

Move session state to Amazon SQS and have the application poll the queue.

Why wrong: SQS is not a session store; it's a message queue.

Network Load Balancer

Why wrong: NLB is for TCP/UDP traffic, not an API gateway.

Application Load Balancer

Why wrong: ALB is a load balancer, not an API gateway.

Amazon CloudFront

Why wrong: CloudFront is a CDN, not an API gateway.

Pass the database credentials as environment variables to the Lambda function.

Why wrong: Passing database credentials as environment variables is insecure because they can be exposed in logs or during function inspection. This does not follow security best practices.

Store the database credentials in AWS Systems Manager Parameter Store and retrieve them in the Lambda function.

Why wrong: AWS Systems Manager Parameter Store can store secrets, but it lacks native automatic rotation and is not the best practice for database credentials requiring rotation. Secrets Manager is preferred.

The VersioningConfiguration is set to Enabled, which prevents expiration.

Why wrong: Versioning does not prevent expiration.

The rule is missing a 'Status' property.

Why wrong: Status is set to Enabled.

The transition to GLACIER storage class prevents expiration from occurring.

Why wrong: Transition does not prevent expiration.

Use Amazon RDS for PostgreSQL with TimescaleDB extension.

Why wrong: Not as scalable as Timestream for IoT scale.

Use Amazon Redshift with automated snapshot retention of 90 days.

Why wrong: Redshift is not low-latency for recent data without streaming.

Use Amazon DynamoDB with TTL to expire data after 90 days and use PartiQL for queries.

Why wrong: PartiQL is limited; DynamoDB is not optimal for time-series analytics.

The object is encrypted with SSE-KMS, not SSE-S3.

Why wrong: The bucket uses SSE-S3.

The user does not have the s3:GetObject permission.

Why wrong: The policy grants s3:GetObject.

The bucket has a bucket policy that denies all requests.

Why wrong: No information about bucket policy.

Take manual snapshots periodically and copy them to us-west-2.

Why wrong: Manual snapshots are taken periodically (e.g., daily) and cannot achieve a 5-minute RPO. Restoring from a snapshot also takes longer than 15 minutes, failing the RTO.

Use Amazon Aurora Global Database with a secondary cluster in us-west-2.

Why wrong: Incorrect. Amazon Aurora Global Database is a feature exclusive to Amazon Aurora, not available for RDS PostgreSQL.

Use automated backups with cross-Region copy to us-west-2.

Why wrong: Incorrect. Automated backups with cross-Region copy provide daily snapshots with recovery times exceeding 15 minutes, thus not meeting the RPO of 5 minutes or RTO of 15 minutes.

Use a NAT Gateway with Elastic IPs for outbound traffic.

Why wrong: Incorrect. A NAT Gateway with Elastic IPs is used for outbound traffic from private subnets, not for providing static IPs for inbound traffic to the ALB.

Use Amazon Route 53 with an A record pointing to the ALB.

Why wrong: Incorrect. Amazon Route 53 with an A record points to the ALB's DNS name, which resolves to dynamic IP addresses that change over time.

Associate an AWS WAF web ACL with the ALB.

Why wrong: Incorrect. AWS WAF is a web application firewall that filters traffic based on rules; it does not provide static IP addresses for the ALB.

Assign the same security group to the ALB and the instances

Why wrong: Assigning the same security group to both ALB and instances is not a best practice and does not enforce encryption or restrict traffic as required.

Place the EC2 instances in a public subnet

Why wrong: Placing EC2 instances in a public subnet would expose them to the internet, violating the requirement that they only accept traffic from the ALB.

Use network ACLs to restrict inbound traffic to the ALB's IP addresses

Why wrong: Network ACLs operate at the subnet level and cannot enforce encryption of traffic between the ALB and EC2 instances; they only filter IP addresses and ports statelessly. This option is tempting because network ACLs are commonly used to restrict inbound traffic by source IP, and limiting to the ALB’s IPs would block non-ALB traffic. However, the requirement for *encrypted* traffic demands a security group referencing the ALB’s security group as source, combined with HTTPS listener configuration, which network ACLs cannot provide.

Use server-side encryption with S3 managed keys (SSE-S3)

Why wrong: Incorrect: SSE-S3 uses AWS-managed keys; the company cannot manage or rotate the keys.

Enable default encryption on the S3 bucket with SSE-S3

Why wrong: Incorrect: Default encryption with SSE-S3 still uses AWS-managed keys, not company-managed.

Use client-side encryption with the Amazon S3 encryption client

Why wrong: Incorrect for this two-answer selection. While client-side encryption does encrypt data at rest with company-managed keys, the exam context often expects server-side encryption options for S3 data lakes. Both SSE-C and SSE-KMS meet the requirements more directly.