SAP-C02 Questions and Answers

D is correct because thehealth check grace perioddefines how long Auto Scaling waits after launching an instance before checking its health status. With the larger content added to the S3 bucket, the instance initialization (via user data) is taking longer. Increasing the grace period allows the instance time to complete startup tasks before it’s marked as unhealthy.

Option A would not necessarily reduce startup time — the issue is likely network latency or the size of the content.

Option B only affects the timeout for health check responses, not the delay before they begin.

Option C is not applicable unless the health check path itself is invalid (which isn ' t mentioned).

https://docs.aws.amazon.com/firehose/latest/dev/creating-the-stream-to-splunk.html

Because MSK has Maximum number of client connections 1000 per second and the company has 10,000 sensors, the MSK likely will not be able to handle all connectionshttps://docs.aws.amazon.com/msk/latest/developerguide/limits.html

migrating the data processing script to an AWS Lambda function and using an S3 event notification to invoke the Lambda function to process the objects when the company uploads the objects. This solution meets the company ' s requirements of high availability and scalability, as well as reducing long-term management overhead, and is likely to be the most cost-effective option.

The requirements specify cross-Region failover with an RPO of 5 minutes and an RTO of 20 minutes for a large (10 TB) database. Achieving a low RPO in a secondary Region typically requires near-continuous replication, not periodic snapshots. Achieving a 20-minute RTO requires that the standby data store already exists in the secondary Region and can be promoted quickly.

A cross-Region read replica provides continuous asynchronous replication from the primary Region to a replica in the secondary Region. This is the standard managed approach for cross-Region disaster recovery for Amazon RDS engines that support read replicas. With a read replica already running in the secondary Region, failover can be accomplished by promoting the replica to a standalone primary database instance. This supports an RTO target like 20 minutes because the data is already present and the promotion operation is typically faster than restoring from a full snapshot, especially at 10 TB scale.

Option B describes creating an RDS Multi-AZ DB cluster in the primary Region and a cross-Region read replica in the secondary Region. Multi-AZ addresses high availability within the primary Region, while the cross-Region replica addresses disaster recovery. The option also includes automation using CloudWatch alarms and a Lambda function to promote the replica when a failure is detected. This supports the required RTO by reducing manual intervention time and supports the RPO by using continuous replication.

Option A is not practical for RPO 5 minutes at 10 TB. Taking and copying snapshots every 5 minutes would be operationally heavy and would not complete within the required window; snapshot creation and cross-Region copy are not designed for such high-frequency near-real-time DR. It also risks missing the RPO due to snapshot completion time.

Option C introduces unnecessary complexity and additional services. Creating a second cluster only at failure time is not compatible with a 20-minute RTO for a 10 TB database, and using DMS for continuous sync is more development and operational work than using native RDS replication. This is not the least overhead or most direct DR design.

Option D is incorrect as stated because cross-Region read replicas are not automatically failed over and promoted by an “automated failover” feature in the way Multi-AZ failover works within a Region. Automated backups do not provide near-real-time cross-Region recovery to meet a 5-minute RPO. The usual approach for cross-Region is to create the replica and then promote it through a controlled process, which can be automated via monitoring and Lambda as in option B.

Therefore, using a cross-Region read replica with automation to promote it provides the necessary RPO/RTO with a managed service approach.

https://docs.aws.amazon.com/AmazonS3/latest/userguide/privatelink-interface-endpoints.html#types-of-vpc-endpoints-for-s3

https://aws.amazon.com/premiumsupport/knowledge-center/s3-bucket-access-direct-connect/

Use a private IP address over Direct Connect (with an interface VPC endpoint)

To access Amazon S3 using a private IP address over Direct Connect, perform the following steps:

3. Create a private virtual interface for your connection.

5. Create an interface VPC endpoint for Amazon S3 in a VPC that is associated with the virtual private gateway. The VGW must connect to a Direct Connect private virtualinterface. This interface VPC endpoint resolves to a private IP address even if you enable a VPC endpoint for S3.

To connect VPC resources over a transit Virtual Interface (VIF) using a Direct Connect connection, the company should advertise the on-premises network prefixes over the transit VIF and advertise the VPC prefixes from the Direct Connect gateway to the on-premises network over the same VIF. This configuration ensures seamless connectivity between the on-premises infrastructure and the AWS VPCs through the transit gateway, facilitating efficient and secure communication across the network.

AWS Documentation on AWS Direct Connect and transit gateways provides detailed instructions on configuring transit VIFs and routing for Direct Connect connections. This setup is recommended in AWS best practices for establishing dedicated network connections between on-premises environments and AWS to achieve low-latency, high-throughput, and secure connectivity.

Configuring the workload to use topology spread constraints that are based on Availability Zone will maximize the node resilience of the workload node groups. This will ensure that the pods are evenly distributed across different Availability Zones, reducing the impact of failures or disruptions in one Availability Zone2. This will also improve the availability and scalability of the workload node groups, as they can leverage the low-latency, high-throughput, and highly redundant networking between Availability Zones1.

Option C is correct because SageMaker AI real-time endpoints support multiple production variants behind a single endpoint. Each production variant can represent a different model version, and traffic distribution is controlled by variant weights. Setting each model variant to 0.5 produces an even split, allowing A/B testing while the existing ecommerce application continues invoking a single endpoint. AWS documentation explains that production variants are used for A/B testing and that traffic to variants is distributed according to configured weights. Option B and option D require application-side routing or round robin logic, which increases change impact and operational complexity. Option A is weaker because the standard documented pattern for controlled A/B testing with weighted model variants is a SageMaker real-time endpoint with production variants.

The company requires two different types of security analysis: vulnerability scanning for CVEs and code-level scanning for sensitive data exposure. Amazon Inspector provides both Lambda Standard Scanning and Lambda Code Scanning. These capabilities allow Inspector to examine Lambda deployment packages and Lambda layers for software vulnerabilities, and also perform static code analysis to detect insecure coding patterns. To meet the requirement that only certain Lambda functions receive deeper code scanning, Inspector allows enabling or disabling code scanning at the individual Lambda function level through the function ' s Monitor settings. Inspector additionally supports resource exclusion based on tagging, allowing administrators to prevent specific Lambda functions from participating in code scans. Tag-based exclusion satisfies the requirement to scan only a subset of Lambda functions.

Option B is required because enabling both Lambda standard scanning and Lambda code scanning activates the functionality necessary for CVE detection and code security analysis.

Option D is required because enabling scanning at the function level ensures that code scanning runs only for selected Lambda functions.

Option E is required because resource exclusion using tags allows the administrator to exclude non-target Lambda functions from code scanning automatically.

Options A and C do not satisfy the requirement because simply activating Amazon Inspector does not configure per-function scanning behavior, and GuardDuty Lambda Protection does not scan for CVEs or code vulnerabilities. Option F is not valid because Amazon Inspector does not perform scans directly against objects stored in Amazon S3; scanning occurs against the deployed Lambda function versions themselves.

Migration Hub Discovery Agentcollectsdeep host-level data, includingprocesses, connections, dependencies, etc. Combined withStrategy Recommendations, it offers accurate insights for planning cloud migration.

Migration Hub Strategy Recommendations

The Auto Scaling group can automatically launch and terminate EC2 instances based on the demand and health of the web application. The launch template can specify the configuration of the EC2 instances, such as the AMI, instance type, security group, and user data. The existing ALB can distribute the traffic to the EC2 instances in the Auto Scaling group. The existing EC2 instances can be attached to the Auto Scaling group without deleting them or the ALB. This option minimizes downtime and preserves the current setup of the web application. References: [What is Amazon EC2 Auto Scaling?] , [Launch templates], [Attach a load balancer to your Auto Scaling group] , [Attach EC2 instances to your Auto Scaling group]

To minimize EKS compute costs:

A. Rebuild the application container images to support ARM64 architecture: Graviton instances (ARM-based) require container images built for the ARM64 architecture. This ensures compatibility and allows leveraging Graviton-based compute.

C. Migrate the EKS nodes to the most recent generation of Graviton-based instances: AWS Graviton instances (e.g., c7g, m7g) offer significant price/performance benefits over x86_64 instances. By migrating EKS worker nodes to Graviton, the company can reduce compute costs substantially.

E. Purchase a new EC2 Instance Savings Plan for the newly selected Graviton instance family: Savings Plans provide cost savings compared to On-Demand pricing. After moving to Graviton, purchasing a new Savings Plan tailored for these instance families ensures continued cost savings.This approach aligns with AWS cost optimization best practices—migrating to the latest instance types and purchasing Savings Plans to match the new usage pattern—while ensuring full compatibility with EKS.

To enhance the network connectivity solution ' s availability, fault tolerance, and security in a cost-effective manner, adding a dynamic private IP AWS Site-to-Site VPN as a secondary path is a viable option. This VPN serves as a resilient backup for the Direct Connect connection, ensuring continuous data flow even if the primary path fails. Implementing MACsec (Media Access Control Security) on the Direct Connect connection further secures the data in transit by providing encryption, thus addressing the security requirement. This solution strikes a balance between cost and operational efficiency, avoiding the higher expenses associated with provisioning an additional Direct Connect connection.

AWS Documentation on AWS Direct Connect and AWS Site-to-Site VPN provides insights into setting up resilient and secure network connections. Additionally, information on MACsec offers guidance on how to implement encryption for Direct Connect connections, aligning with best practices for secure and highly available network architectures.

https://aws.amazon.com/blogs/networking-and-content-delivery/deployment-models-for-aws-network-firewall/

They all mention using spot instances and EKS based on EC2. A spot instance is not appropriate for a production server and the company is developing new application designed for AWS Fargate, which means we must plan the future cost improvement including AWS Fargate. https://aws.amazon.com/savingsplans/compute-pricing/

The company wants a centralized enforcement mechanism across hundreds of AWS accounts. The control must ensure that any S3 access points that product teams create are VPC-only and cannot be internet-accessible. The most operationally efficient solution is one that applies across the organization without requiring per-account deployments, per-bucket policies, or per-access-point configuration.

Service control policies (SCPs) in AWS Organizations are designed to provide centralized guardrails that define the maximum available permissions for accounts in an organization. By attaching an SCP at the organization root (or to OUs as needed), the company can enforce that no principal in any account can create an S3 access point unless the request specifies a VPC network origin. This aligns directly with the requirement to enforce “VPC-only” access points consistently across all accounts with minimal ongoing operational work.

Option B uses an SCP at the root to deny s3:CreateAccessPoint unless the s3:AccessPointNetworkOrigin condition key evaluates to VPC. This is the correct organization-wide preventive control.

Option A is not correct because an S3 access point resource policy is attached to an access point after it exists and is primarily used to control access to the access point. It is not a reliable organization-wide preventive mechanism to enforce how access points are created across hundreds of accounts. Also, it requires access point-by-access point management, which increases operational overhead.

Option C is less operationally efficient because StackSets deployment across hundreds of accounts introduces additional operational steps and drift management. It also relies on account-level IAM permissions being properly used and does not provide a simple, centralized “cannot be bypassed” guardrail in the same way that an SCP does.

Option D is incorrect because S3 bucket policies control access to bucket resources and objects. They do not function as an organization-wide control to prevent creation of access points across accounts, and they would require bucket-by-bucket management rather than a centralized enforcement mechanism.

Therefore, a root-level SCP that denies access point creation unless the access point network origin is VPC is the most operationally efficient enforcement approach.

https://joe.blog.freemansoft.com/2020/04/protect-data-in-cloud-with-s3-access.html

Converting VPC flow logs to store in Apache Parquet format and specifying hourly partitions significantly improves query performance and reduces storage space usage. Apache Parquet is a columnar storage file format optimized for analytical queries, allowing Athena to scan less data and improve query performance. Partitioning logs by hour further enhances query efficiency by limiting the amount of data scanned during queries, addressing the issue of degrading performance over time due to the growing volume of ingested logs.

AWS Documentation on VPC Flow Logs and Amazon Athena provides insights into configuring VPC flow logs in Apache Parquet format and using Athena for querying log data. This approach is recommended for efficient log analysis and storage optimization.

Option A is correct because AWS Compute Optimizer provides right-sizing recommendations for Amazon ECS services on AWS Fargate, including task CPU, task memory, and container-level CPU and memory values. Since the workload is deployed through CloudFormation, the correct operational path is to update the ECS task definition in the CloudFormation template and redeploy the stack. That keeps infrastructure as code as the source of truth. Target tracking policies adjust the number of running tasks, not the CPU and memory size assigned to each Fargate task, so option B does not directly fix over-provisioned task resources. Cost Explorer Reserved Instance reports are not the right tool for Fargate task right-sizing because Fargate does not use EC2 Reserved Instances in the same way.

S3 Event Notifications is a feature that allows users to receive notifications when certain events happen in an S3 bucket, such as object creation or deletion1. Users can configure S3 Event Notifications to invoke an AWS Lambda function when a user stores an image in the bucket. Lambda is a serverless compute service that runs code in response to events and automatically manages the underlying compute resources2. The Lambda function can resize the image in place and store the original file in the same S3 bucket. This way, the solution can handle significant variance in demand and be reliable at enterprise scale. Thesolution can also rerun processing jobs in the event of failure by using the retry and dead-letter queue features of Lambda2.

S3 Lifecycle is a feature that allows users to manage their objects so that they are stored cost-effectively throughout their lifecycle3. Users can create an S3 Lifecycle policy to move all stored images to S3 Standard-Infrequent Access (S3 Standard-IA) after 6 months. S3 Standard-IA is a storage class designed for data that is accessed less frequently, but requires rapid access when needed4. It offers a lower storage cost than S3 Standard, but charges a retrieval fee. Therefore, moving the images to S3 Standard-IA after 6 months can reduce the storage cost for the solution.

Option A is incorrect because using AWS Step Functions to process the S3 event that occurs when a user stores an image is not necessary or cost-effective. AWS Step Functions is a service that lets users coordinate multiple AWS services into serverless workflows. However, for this use case, a single Lambda function can handle the image resizing task without needing Step Functions.

Option B is incorrect because using Amazon EventBridge to process the S3 event that occurs when a user uploads an image is not necessary or cost-effective. Amazon EventBridge is a serverless event bus service that makes it easy to connect applications with data from a variety of sources. However, for this use case, S3 Event Notifications can directly invoke the Lambda function without needing EventBridge.

Option D is incorrect because using Amazon Simple Queue Service (Amazon SQS) to process the S3 event that occurs when a user stores an image is not necessary or cost-effective. Amazon SQS is a fully managed message queuing service that enables users to decouple and scale microservices, distributed systems, and serverless applications. However, for this use case, S3 Event Notifications can directly invoke the Lambda function without needing SQS. Moreover, storing the resized file in an S3 bucket that uses S3 Standard-IA will incur a retrieval fee every time the file is accessed, which may not be cost-effective for frequently accessed files.

https://aws.amazon.com/blogs/database/cost-effective-disaster-recovery-for-amazon-aurora-databases-using-aws-backup/

Migrating the containers that run the application to AWS Fargate for Amazon Elastic Container Service (Amazon ECS) will meet the requirement of not administering the Docker hosting environment. AWS Fargate is a serverless compute engine that runs containerswithout requiring any infrastructure management3. Creating an Amazon Elastic File System (Amazon EFS) file system and adding a volume to the Fargate task definition to point to the EFS file system will meet the requirement of low-latency storage that will automatically scale throughput to meet increased demand. Amazon EFS is a fully managed file system service that provides shared access to data from multiple containers, supports NFSv4 protocol, and offers consistent performance and high availability4. Amazon EFS also supports automatic scaling of throughput based on the amount of data stored in the file system5.

https://aws.amazon.com/solutions/implementations/disaster-recovery-for-aws-iot/

Option D is correct because Amazon ECR enhanced scanning integrates with Amazon Inspector and provides continuous vulnerability scanning for container images. Enhanced scanning covers operating system package vulnerabilities and programming language package vulnerabilities, which basic scanning does not fully satisfy. Amazon Inspector can also help prioritize remediation by showing where vulnerable images are actively deployed, such as in Amazon ECS or Amazon EKS environments. EventBridge integration allows findings to be routed into an existing event-driven workflow without custom scanning infrastructure. Option A uses basic scanning, which is insufficient for the stated language package and continuous monitoring requirements. Option B creates unnecessary custom scanning and storage logic. Option C incorrectly combines basic and enhanced scanning with manual triggers, which does not meet the least-operational-overhead requirement.

The company needs the application in the secondary Region to operate independently from the primary Region during disaster recovery. That means both the DynamoDB data and the relational customer data must be present and usable in the secondary Region without cross-Region dependencies at runtime. The company also wants the least development effort, which usually means using native managed replication features rather than building custom replication logic.

For DynamoDB, global tables are the managed multi-Region, multi-active replication feature. Global tables replicate changes between Regions automatically and allow applications in each Region to read and write to local DynamoDB tables, removing dependencies on the primary Region.

For the RDS for MySQL database, the straightforward managed approach for cross-Region disaster recovery with minimal development change is to create a cross-Region read replica in the secondary Region. A read replica uses asynchronous replication to keep data up to date. During a disaster recovery event, the company can promote the read replica in the secondary Region to become a standalone primary database instance. This provides a simple, low-development approach for having the data in the secondary Region. The application in the secondary Region can be configured to use the replica endpoint in normal times and can continue operating with the promoted instance during failover.

Option A combines DynamoDB global tables with an RDS read replica in the secondary Region. This is the least development effort because it uses AWS-managed replication for both data stores and requires only configuration changes and endpoint selection in the secondary application.

Option B is not correct because DAX is a caching layer for DynamoDB, not a replication or disaster recovery mechanism. DAX does not replicate DynamoDB tables into another Region and cannot replace having the DynamoDB data present in the secondary Region.

Option C is incorrect because RDS Multi-AZ is a high availability feature within a single Region, using synchronous replication to a standby in another Availability Zone. Multi-AZ does not create a standby in a different Region, and it does not provide cross-Region DR by itself.

Option D is less desirable because it requires custom development and operations: processing DynamoDB streams, building and operating a replication pipeline, and ensuring correctness and ordering across Regions. This is higher development effort than enabling DynamoDB global tables.

Therefore, using DynamoDB global tables plus an RDS read replica in the secondary Region (option A) meets the DR independence requirement with the least development effort.

The core issue is high latency when serving images during traffic spikes. The current design resizes images dynamically and caches them locally inside ECS containers. With AWS Fargate, containers are ephemeral and scale in and out based on load. Local container caches are not shared across tasks, so cache hit rates drop during scaling events, forcing repeated image resizing. This increases CPU usage, startup latency, and response times under load.

The lowest operational overhead and most effective solution is to offload image delivery entirely from the application layer and use edge caching. Amazon CloudFront is designed to cache static content at edge locations close to end users, dramatically reducing latency and backend load. Amazon S3 is a durable, highly scalable object store that is well suited for serving static assets such as images.

Option A moves image delivery to a CloudFront + S3 architecture. By pre-scaling images and storing them in S3, the application no longer performs on-demand image resizing during requests. CloudFront caches the images at edge locations, so subsequent requests are served directly from the edge with minimal latency. The ALB remains the origin for dynamic content, while image requests are routed to the S3 origin using cache behaviors. This approach is fully managed, highly scalable, and requires minimal changes to the application logic once images are pre-generated.

Option B introduces ElastiCache, which adds operational overhead and complexity. The application must manage cache keys, eviction, and failure scenarios. ElastiCache is better suited for low-latency data access patterns rather than large static objects like images, and it does not provide global edge caching.

Option C adds unnecessary complexity by introducing an Aurora database as part of the caching layer. Databases are not designed to manage cache metadata for static assets at scale, and this design increases operational overhead without improving latency compared to CloudFront.

Option D is not appropriate because API Gateway caching is intended for API responses, not for efficient delivery of large static binary objects such as images. Replacing the ALB with API Gateway also introduces request size, payload, and cost considerations and does not provide global edge caching comparable to CloudFront for image delivery.

Therefore, using Amazon CloudFront with Amazon S3 for pre-scaled images provides the lowest latency, highest scalability, and least operational overhead.

Creating an internal ALB and configuring it as a PrivateLink endpoint service enables private connectivity between internal applications and the metadata API, ensuring that traffic does not traverse the public internet.

Internal ALB: Ensures traffic stays within the AWS network and is not exposed publicly.

PrivateLink endpoint service: Provides secure, private access to the ALB from the internal EC2 instances in other AWS accounts.

Traffic stays within the AWS global network, leveraging AWS security best practices and meeting the new policy requirements for no public internet exposure.This approach is secure, scalable, and minimizes management complexity compared to API Gateway solutions.

https://aws.amazon.com/about-aws/whats-new/2017/11/announcing-support-for-inter-region-vpc-peering/

The company wants backups to be stored in an isolated central account designed for WORM storage. With AWS Organizations, AWS Backup supports centralized, cross-account backup management through delegated administration, backup policies, and cross-account backup vault usage. This provides the least operational overhead and scales to dozens of databases.

A key constraint is encryption. The databases are encrypted with the default RDS AWS KMS key. AWS-managed KMS keys (including the default RDS KMS key) are not customer-managed and generally cannot be shared across accounts for cross-account snapshot copy and cross-account backup workflows that require key sharing and explicit grants. For cross-account backup copies, the source backups must be encrypted with a customer managed KMS key that can be shared (via key policy and grants) with the destination account. Therefore, the company must move away from using the default RDS AWS-managed key for encryption of snapshots that will be copied across accounts.

Option A uses the correct managed approach: enable trusted access and backup policies in Organizations, make the central account the delegated administrator, and use AWS Backup to manage cross-account backups into a central backup vault. It also correctly introduces a customer managed KMS key in the central account for the backup vault and shares that key with the production account. Because the existing databases are encrypted with the default RDS AWS-managed key, option A includes the necessary remediation step: restore (or otherwise recreate) the databases so they are encrypted with the shared customer managed KMS key, enabling future backups and copies to be encrypted with a shareable key and stored in the central vault.

Option B is not workable because it requires “sharing the default RDS KMS key” with another account and using it for the central vault. AWS-managed keys are not shared and are not controlled by customers in a way that supports this cross-account backup model. Therefore, this option fails the encryption requirement.

Option C and option D rely on decrypting snapshots, copying them as unencrypted, and then re-encrypting. This is not a valid or safe operational pattern for RDS automated backups at scale, and it conflicts with typical constraints around snapshot encryption transitions. It also introduces substantial custom automation and operational overhead, and it undermines the objective of a controlled WORM backup account by manipulating encryption state through custom code. Additionally, “decrypting snapshots” is not how encrypted RDS snapshots are typically handled; encrypted snapshots remain encrypted and are copied re-encrypted with a different KMS key that the destination can use, which requires a customer managed key and proper permissions.

Therefore, the successful and scalable solution is to centralize backups with AWS Backup and ensure the databases are encrypted with a customer managed KMS key that can be shared cross-account, as described in option A.

https://repost.aws/knowledge-center/vpc-ip-address-range

DynamoDB with TTL, cheaper for sustained throughput of small items + suited for fast retrievals. S3 cheaper for storage only, much higher costs with writes. RDS not designed for this use case.

The best solution is to upload files from the mobile software directly to Amazon S3, use S3 event notifications to create a message in an Amazon Simple Queue Service (Amazon SQS) standard queue, and invoke an AWS Lambda function to perform image processing when a message is available in the queue. This solution will ensure that image processing can scale to handle the load, as Amazon S3 can store any amount of data and handle concurrent uploads, Amazon SQS can buffer the messages and deliver them reliably, and AWS Lambda can run code without provisioning or managing servers and scale automatically based on the demand. This solution will also notify the user when processing is complete by sending a push notification to the mobile app using Amazon Simple Notification Service (Amazon SNS), which is a web service that enables applications to send and receive notifications from the cloud. This solution is more cost-effective than using Amazon MQ, which is a managed message broker service for Apache ActiveMQ that requires a dedicated broker instance, or S3 Batch Operations, which is a feature that allows users to perform bulk actions on S3 objects, such as copying or tagging, but does not support custom code execution. This solution is also more suitable than using Amazon Simple Email Service (Amazon SES), which is a web service that enables applications to send and receive email messages, but does not support push notifications for mobile devices. References: Amazon S3 Documentation, Amazon SQS Documentation, AWS Lambda Documentation, Amazon SNS Documentation

to connect out from the private subnet you need an NAT gateway and since only one Elastic IP whitelisted on firewall its one NATGateway at time and if AZ failure happens Lambda creates a new NATGATEWAY in a different AZ using the Same Elastic IP ,dont be tempted to select D since application that needs to connect is on a private subnet whose outbound connections use the NATGateway Elastic IP

Comprehensive and Detailed Explanation:

A. Using AWS CloudFormation templates allows for the automation of infrastructure deployment. By parameterizing the templates, resources can be provisioned consistently across multiple Regions, reducing administrative overhead and ensuring infrastructure as code practices.

D. Amazon Route 53 latency-based routing directs user requests to the AWS Region that provides the lowest latency, improving access times for users globally and enhancing the user experience.

E. Enabling DynamoDB Streams and adding a second Region to create a global table ensures that data is replicated across Regions, providing high availability and disaster recovery capabilities.

This combination of steps ensures that the application stack is replicated efficiently across Regions, providing disaster recovery, improved performance, and scalability, all while minimizing manual administrative tasks.

The analytics team wants to copy only a subset of the data, run the copy as soon as data is ready, and receive notifications when the process completes. AWS DataSync supports the use of manifest files to control exactly which files or objects are transferred. By generating a manifest file on premises that lists only the relevant data and uploading that manifest to Amazon S3, the company can precisely limit what DataSync copies, which reduces data transfer costs and processing overhead.

Option A satisfies the requirement to identify only analytics-relevant data by creating and uploading a manifest file that lists the objects to be transferred. This avoids copying unnecessary data and is cost-effective.

Option D uses Amazon S3 Event Notifications to trigger an AWS Lambda function as soon as the manifest file is uploaded. The Lambda function starts the DataSync task and references the manifest file. This ensures that the copy process begins immediately after data generation, rather than waiting for a fixed schedule, which meets the “as soon as possible” requirement.

Option E provides a fully managed and scalable notification mechanism. AWS DataSync publishes task execution state changes as events. Amazon EventBridge can capture SUCCESS or ERROR states and route those events to Amazon SNS, which can then send email notifications. This approach avoids custom polling logic and minimizes operational overhead.

Option B and C introduce Amazon DynamoDB unnecessarily. DataSync can directly consume a manifest file from Amazon S3, so loading manifest data into DynamoDB and orchestrating copy logic manually adds cost and complexity without benefit.

Option F relies on a custom Lambda function to monitor task state changes, which increases operational overhead compared to the native EventBridge integration with DataSync.

Therefore, using S3-based manifests, event-driven invocation of DataSync, and EventBridge-based notifications is the most cost-effective and operationally efficient solution.

The AWS Application Discovery Service can help plan migration projects by collecting data about on-premises servers, such as configuration, performance, and network connections. The data collection agent is a lightweight software that can be installed on each server to gather this information. This option is more cost-effective than agentless discovery, which requires deploying a virtual appliance in the VMware environment, or using CloudWatch agent, which incurs additional charges for CloudWatch Logs. Scanning the servers over a VPN is not a valid option for AWS Application Discovery Service. References: What is AWS Application Discovery Service?, Data collection methods

AWS Elastic Disaster Recovery (AWS DRS) is a service that minimizes downtime and data loss with fast, reliable recovery of on-premises and cloud-based applications using affordable storage, minimal compute, and point-in-time recovery1. Users can set up AWS DRS on their source servers to initiate secure data replication to a staging area subnet in their AWS account, in the AWS Region they select. Users can then launch recovery instances on AWS within minutes, using the most up-to-date server state or a previous point in time.

To configure a cloud backup of the application with AWS DRS, users need to create a VPC that has at least two public subnets, a virtual private gateway, and an internet gateway. AVPC is a logically isolated section of the AWS Cloud where users can launch AWS resources in a virtual network that they define2. A public subnet is a subnet that has a route to an internet gateway3. A virtualprivate gateway is the VPN concentrator on the Amazon side of the Site-to-Site VPN connection4. An internet gateway is a horizontally scaled, redundant, and highly available VPC component that allows communication between instances in the VPC and the internet. Users need to create at least two public subnets for redundancy and high availability. Users need to create a virtual private gateway and attach it to the VPC to enable VPN connectivity between the on-premises network and the target AWS network. Users need to create an internet gateway and attach it to the VPC to enable internet access for the replication servers.

To ensure that replication traffic does not travel through the public internet, users need to create an AWS Direct Connect connection and a Direct Connect gateway between the on-premises network and the target AWS network. AWS Direct Connect is a service that establishes a dedicated network connection from an on-premises network to one or more VPCs. A Direct Connect gateway is a globally available resource that allows users to connect multiple VPCs across different Regions to their on-premises networks using one or more Direct Connect connections. Users need to create an AWS Direct Connect connection between their on-premises network and an AWS Region. Users need to create a Direct Connect gateway and associate it with their VPC and their Direct Connect connection.

To ensure that the application is not accessible from the internet, users need to select the option to use private IP addresses for data replication during configuration of the replication servers. This option configures the replication servers with private IP addresses only, without assigning any public IP addresses or Elastic IP addresses. This way, the replication servers can only communicate with other resources within the VPC or through VPN connections.

Option A is incorrect because creating a VPC that has at least two private subnets, two NAT gateways, and a virtual private gateway is not necessary or cost-effective. A private subnet is a subnet that does not have a route to an internet gateway3. A NAT gateway is a highly available, managed Network Address Translation (NAT) service that enables instances in a private subnet to connect to the internet or other AWS services, but prevents the internet from initiating connections with those instances. Users do not need to createprivate subnets or NAT gateways for this use case, as they can use public subnets with private IP addresses for data replication.

Option C is incorrect because creating an AWS Site-to-Site VPN connection between the on-premises network and the target AWS network will not ensure that replication traffic does not travel through the public internet. A Site-to-Site VPN connection consists of two VPN tunnels between an on-premises customer gateway device and a virtual private gateway in your VPC4. The VPN tunnels are encrypted using IPSec protocols, but they still use public IP addresses for communication. Users need to use AWS Direct Connect instead of Site-to-Site VPN for this use case.

Option F is incorrect because selecting the option to ensure that the Recovery instance’s private IP address matches the source server’s private IP address during configuration of the launch settings for the target servers will not ensure that the application is not accessible from the internet. This option configures the Recovery instance with an identical private IP address as its source server when launched in drills or recovery mode. However, this option does not prevent assigning public IP addresses or Elastic IP addresses to the Recovery instance. Users need to select the option to use private IP addresses for data replication instead.

By splitting the 12 instances across three Availability Zones, the system can maintain high availability and availability of resources in case of a failure. Option D also uses a combination of On-Demand Instances with Capacity Reservations and Spot Instances, which allows for scheduled jobs to be run on the On-Demand instances with guaranteed capacity, while also taking advantage of the cost savings from Spot Instances for the user jobs which have lower SLA requirements.

https://docs.aws.amazon.com/controltower/latest/userguide/controls.html https://docs.aws.amazon.com/controltower/latest/userguide/strongly-recommended-controls.html#ebs-enable-encryption AWS is now transitioning the previous term ' guardrail ' new term ' control ' .

The company wants logging that helps troubleshoot email delivery issues and also wants to search by recipient, subject, and time sent. Amazon SES provides event publishing for email sending and delivery events through configuration sets. With a configuration set, SES can publish sending events such as send, delivery, bounce, complaint, reject, and rendering failure to destinations such as Amazon CloudWatch, Amazon SNS, or Amazon Kinesis Data Firehose. For building a searchable log store, delivering these events into Amazon S3 through Kinesis Data Firehose is an effective approach because S3 provides durable storage and integrates well with query services.

Option A creates an SES configuration set with a Kinesis Data Firehose destination that delivers logs to an S3 bucket. This captures detailed SES event data that is directly useful for troubleshooting delivery issues and retaining historical records for analysis.

Once logs are stored in Amazon S3, Amazon Athena can query the data using SQL. Athena is designed to query data in S3 and is well suited for ad hoc searches. This meets the requirement to search based on recipient, subject, and time sent, assuming the event schema includes these fields (or they are included in the published event payload). Therefore, option C completes the solution by enabling searches over the stored log dataset.

Option B (CloudTrail) records API activity, such as calls made to SES APIs, but it is not designed to capture per-message delivery outcomes (deliveries, bounces, complaints) in a way that supports troubleshooting delivery behavior and detailed email event searching. CloudTrail is useful for auditing who called SES APIs, not for tracking message-level delivery events and outcomes.

Option D (CloudWatch log group) is another valid SES event publishing destination, but if the requirement is to perform flexible searches by multiple dimensions over a potentially large historical dataset, storing the logs in S3 and querying with Athena is a more direct and scalable pattern. Also, the provided option E is incorrect because Athena queries data in S3, not in CloudWatch Logs. CloudWatch Logs has its own query mechanism, but Athena is not used to query CloudWatch Logs directly in the way described.

Option E is incorrect because Amazon Athena does not query CloudWatch Logs as a log store. The typical searchable pattern for Athena is S3-backed datasets.

Therefore, the best combination to satisfy logging and searchable analysis requirements is to publish SES events to S3 via Kinesis Data Firehose (option A) and query those logs with Athena (option C).

Developing infrastructure services using AWS CloudFormation templates and uploading them as AWS Service Catalog products to portfolios created in a central AWS account will enable thecompany to centrally manage the creation of infrastructure services and control who can use them1. AWS Service Catalog allows you to create and manage catalogs of IT services that are approved for use on AWS2. You can organize products into portfolios, which are collections of products along with configuration information3. You can share portfolios with other accounts in your organization using AWS Organizations4.

Allowing user IAM roles to have ServiceCatalogEndUserAccess permissions only and using an automation script to import the central portfolios to local AWS accounts, copy the TagOption, assign users access, and apply launch constraints will enable the company to provide least privilege access to users when launching AWS infrastructure services. ServiceCatalogEndUserAccess is a managed IAM policy that grants users permission to list and view products and launch product instances. An automation script can help import the shared portfolios from the central account to the local accounts, copy the TagOption from the central account, assign users access to the portfolios, and apply launch constraints that specify which IAM role or user can provision a product.

Using the AWS Service Catalog TagOption Library to maintain a list of tags required by the company and applying the TagOption to AWS Service Catalog products or portfolios will enable the company to enforce tags on any infrastructure that is started by users. TagOptions are key-value pairs that you can use to classify your AWS Service Catalog resources. You can create a TagOption Library that contains all the tags that you want to use across your organization. You can apply TagOptions to products or portfolios, and they will be automatically applied to any provisioned product instances.

Creating a product from an existing CloudFormation template

What is AWS Service Catalog?

Working with portfolios

Sharing a portfolio with AWS Organizations

[Providing least privilege access for users]

[AWS managed policies for job functions]

[Importing shared portfolios]

[Enforcing tag policies]

[Working with TagOptions]

[Creating a TagOption Library]

[Applying TagOptions]

This option allows the company to use DAX to improve the performance and reduce the latency of the DynamoDB queries by caching the results in memory1. By updating the application to use DAX, the company can reduce the load on the DynamoDB tables and avoid throttling errors1. By creating an Auto Scaling group for the EC2 instances, the company can adjust the number of instances based on the demand and ensure high availability2. By creating an ALB, the company can distribute the incoming traffic across multiple EC2 instances and improve fault tolerance3. By updating the Route 53 record to use a simple routing policy that targets the ALB’s DNS alias, the company can route users to the ALB endpoint and leverage its health checks and load balancing features4. By configuring scheduled scaling for the EC2 instances before the content updates, the company can anticipate and handle traffic spikes during peak periods5.

What is Amazon DynamoDB Accelerator (DAX)?

What is Amazon EC2 Auto Scaling?

What is an Application Load Balancer?

Choosing a routing policy

Scheduled scaling for Amazon EC2 Auto Scaling

https://aws.amazon.com/premiumsupport/knowledge-center/aws-sftp-endpoint-type/

For migration planning, AWS provides specific tooling to discover on-premises servers, collect detailed performance and configuration data, and analyze application dependencies. The primary service for this purpose is AWS Application Discovery Service, which integrates with AWS Migration Hub.

The AWS Application Discovery Agent is installed on on-premises physical servers and VMs. It collects detailed information such as CPU, memory, and disk utilization; processes and services running on the system; system configuration details; and network connections between systems. This allows the company to understand how servers communicate, what applications are running, and what dependencies exist between components.

This data is sent to AWS Application Discovery Service and surfaced through AWS Migration Hub, where the company can view discovered servers, group them logically into applications, and analyze migration readiness. Migration Hub can use this collected data to provide recommendations for EC2 instance sizing that are based on actual on-premises utilization metrics, helping choose appropriate EC2 instance types and sizes.

Option C directly describes installing the AWS Application Discovery Agent, using AWS Migration Hub to discover and group servers into applications, and using Migration Hub to generate EC2 instance recommendations. This aligns with AWS’s migration assessment tooling and meets all the stated goals: performance measurement, system configuration listing, network connection understanding, dependency analysis, and instance sizing recommendations.

Option A uses AWS Systems Manager Agent and Application Manager. While Systems Manager can manage servers and collect some information, it is not the primary tool for detailed migration discovery, networking dependency mapping, and migration-focused analysis. It also suggests using AWS Pricing Calculator for instance recommendations, which requires manual input and does not automatically derive recommendations from observed utilization and dependencies.

Option B suggests using the Amazon Inspector agent to gather performance and usage information. Amazon Inspector is designed for vulnerability and security assessments, not for detailed migration discovery or dependency mapping. Additionally, AWS Compute Optimizer currently bases its recommendations on usage metrics from AWS resources, not on-premises servers.

Option D uses the CloudWatch agent to collect metrics and then Migration Hub and Compute Optimizer. The CloudWatch agent is not the standard agent for migration discovery and dependency analysis of on-premises workloads. Also, Compute Optimizer focuses on optimization of existing AWS resources rather than creating recommendations based on on-premises utilization.

Therefore, installing the AWS Application Discovery Agent and using AWS Migration Hub (option C) is the correct solution to satisfy all the requirements for migration assessment and EC2 sizing recommendations.

Option D offers the lowest latency and highest efficiency:

Amazon DynamoDB global tables replicate data across multiple Regions, ensuring that the database is available and responsive in each Region.

Lambda@Edge allows the backend logic to execute at AWS edge locations, reducing the distance between the user ' s device and the code execution environment, thereby minimizing latency.

Amazon CloudFront serves as the content delivery network, routing requests to the nearest edge location where the Lambda@Edge function is executed.

Amazon Route 53 maps the DNS name to the CloudFront distribution, ensuring that users are directed to the optimal edge location.

This architecture ensures rapid response times and high availability for a globally distributed user base.

You can use the management account of the organization in AWS Billing and Cost Management console to turn off RI sharing for the HR department ' s production AWS account. This will prevent other departments from sharing the RI discounts and ensure that only the HR department can use the RIs purchased in their production account.

Deploy DataSync Agent:

Install the AWS DataSync agent as a VM in your Hyper-V environment. This agent facilitates the data transfer between your on-premises storage and AWS.

Configure Source and Destination:

Set up the source location to point to your on-premises NFS server where the image batches are stored.

Configure the destination location to be an Amazon S3 bucket with the Glacier Deep Archive storage class. This storage class is cost-effective for long-term storage with retrieval times of up to 12 hours.

Create DataSync Tasks:

Create and configure DataSync tasks to manage the data transfer. Schedule these tasks to run during non-business hours to minimize bandwidth usage during peak times. The tasks will handle the copying of data batches from the NFS server to the S3 bucket.

Set Bandwidth Limits:

In the DataSync configuration, set bandwidth limits to control the amount of data being transferred at any given time. This ensures that your network ' s performance is not adversely affected during business hours.

Delete On-Premises Data:

After successfully copying the data to S3 Glacier Deep Archive, configure the DataSync task to delete the data from your on-premises NFS server. This helps manage storage capacity on-premises and ensures data is securely archived on AWS.

This approach leverages AWS DataSync for efficient, secure, and automated data transfer, and S3 Glacier Deep Archive for cost-effective long-term storage.

References

AWS DataSync Overview【41】.

AWS Storage Blog on DataSync Migration【40】.

Amazon S3 Transfer Acceleration Documentation【42】.

The correct approach involves activating the BusinessUnit cost allocation tag in the management account to ensure costs are tagged consistently across the organization.

The AWS Cost and Usage Report (CUR), delivered to a single Amazon S3 bucket in the management account, consolidates cost and usage data across all member accounts.

Amazon Athena can efficiently query the CUR data directly from S3, while Amazon QuickSight can be used to create interactive dashboards and reports that visualize the cost allocation by business unit.

This solution minimizes complexity, centralizes reporting, and leverages AWS native services for the best cost optimization and insights.

The company has migrated an application to EC2 and will deploy it in an additional Region. The usage is expected to be stable for 3 years, but parts of the application will be redesigned over the next 2 years to use AWS Lambda. Therefore, the company will gradually shift a portion of its compute usage from EC2 to Lambda.

Savings Plans are a flexible pricing model that provide lower prices in exchange for a commitment to a consistent amount of compute usage (measured in dollars per hour) for a 1- or 3-year term. There are two main types of Savings Plans:

• Compute Savings Plans: Apply to any EC2 instance usage regardless of Region, instance family, OS, tenancy, and also apply to AWS Fargate and AWS Lambda usage.

• EC2 Instance Savings Plans: Apply only to a specific instance family in a specific Region.

In this scenario, because there will be a gradual migration from EC2-based compute to Lambda-based compute, a plan that only applies to EC2 (EC2 Instance Savings Plan) risks underutilization as more usage moves to Lambda. Compute Savings Plans, by contrast, can cover EC2 now and Lambda later, maintaining high utilization of the commitment and maximizing effective discount.

Option A recommends evaluating Savings Plans recommendations each year in AWS Cost Management and purchasing a 1-year Compute Savings Plan based on those recommendations. The Cost Management tools and Cost Explorer provide Savings Plans recommendations based on actual usage patterns. Buying 1-year Compute Savings Plans and re-evaluating annually allows the company to adjust its committed spend each year as portions of the application move from EC2 to Lambda. This approach maximizes the chance that the Savings Plans coverage matches the actual combined EC2 and Lambda usage over time and avoids being locked into an EC2-only commitment that becomes increasingly underutilized.

Option B is not optimal because it recommends a 3-year EC2 Instance Savings Plan. That commitment applies only to specific EC2 instance families in a Region and does not apply to Lambda. As the company migrates portions of its workloads to Lambda, the EC2 Instance Savings Plan may become underutilized, reducing the effective discount and potentially increasing overall cost. Compute Optimizer also does not provide Savings Plans purchase recommendations; Savings Plans recommendations come from Cost Management and Cost Explorer.

Option C uses 1-year EC2 Instance Savings Plans and adjusts the commitment each year. While this gives some flexibility, EC2 Instance Savings Plans still do not apply to Lambda usage, so over time there is a risk of underutilizing the EC2 commitment as the Lambda portion grows. This does not maximize savings relative to Compute Savings Plans, which can cover both services.

Option D proposes a 3-year EC2 Instance Savings Plan and suggests decreasing the hourly commitment during the term as workloads move to Lambda. However, Savings Plans commitments cannot be decreased during the term. The company would remain obligated to the original 3-year commitment even if EC2 usage drops, leading to wasted commitment.

Therefore, purchasing 1-year Compute Savings Plans each year based on Savings Plans recommendations from AWS Cost Management (option A) is the best strategy to maximize cost savings while the company transitions part of the workload from EC2 to Lambda.

https://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/lambda-edge-how-it-works-tutorial.html

The best solution is to provision a new Amazon Elastic File System (Amazon EFS) file system that uses Max I/O performance mode and mount the EFS file system on each EC2 instance in the cluster. Amazon EFS is a fully managed, scalable, and elastic file storage service that supports the POSIX standard and can be accessed by multiple EC2 instances concurrently. Amazon EFS offers two performance modes: General Purpose and Max I/O. Max I/O mode is designed for highly parallelized workloads that can tolerate higher latencies than the General Purpose mode. Max I/O mode provides higher levels of aggregate throughput and operations per second, which are suitable for big data analytics applications. This solution meets all the requirements of the company. References: Amazon EFS Documentation, Amazon EFS performance modes

Option C is correct because AWS Systems Manager Patch Manager is the AWS-native service for scalable patch management across managed instances. Patch baselines define approved patches, while delegated administration and OU targeting allow centralized control across many accounts in AWS Organizations. Applying patches first to the non- production OU and then to the production OU with approval gates satisfies the security policy and reduces rollout risk. Systems Manager Compliance and AWS CloudTrail provide auditability for patch status and administrative actions. Option A relies heavily on custom Automation and EventBridge logic, which increases overhead. Option B requires account-by-account configuration and does not scale cleanly across thousands of instances. Option D uses AWS Config for detection, but Config does not provide full patch orchestration by itself.

C is correct because it converts the single–Availability Zone, single-EC2-instance ECS design into a managed, multi-AZ, self-healing architecture with minimal day-to-day operations. The current design has multiple single points of failure: one EC2 instance for ECS capacity and one Availability Zone for all components. Moving the ECS service to AWS Fargate removes the need to manage EC2 instances (capacity provisioning, patching, and scaling of the container instances) and allows the service to run tasks across multiple Availability Zones for higher availability. On the database side, modifying Aurora PostgreSQL to a Multi-AZ DB cluster (by adding a replica in another AZ) increases availability and supports faster recovery from an AZ failure with AWS-managed failover.

Why the other options are less suitable:

A: Making Aurora Multi-AZ improves database availability, but it does not address the compute layer’s biggest issue: ECS is on a single EC2 instance in one AZ with no auto scaling. RDS Proxy can help with connection management, but it does not fix the application’s single-AZ ECS single-instance availability risk.

B: Cross-Region read replicas and manual failover scripts increase operational burden. Also, it keeps ECS on EC2 (still requires instance management) and introduces a manual failover process, which is the opposite of “least operational overhead.”

D: Multi-Region active-active plus Aurora global database can deliver very high availability, but it adds significant complexity (multi-Region deployment, routing strategy, global database considerations, operational procedures). That is higher operational overhead than a straightforward multi-AZ design using managed services.

The best solution is to create an AWS WAF web ACL and configure a rule to block any requests that do not originate from the specified country. This will ensure that only users from the allowed country can access the application. AWS WAF also provides logging capabilities that can capture the access requests that have been blocked. This solution requires the least possible maintenance as it does not involve updating IP ranges or security group rules. References: [AWS WAF Developer Guide], [AWS Shield Developer Guide]

The best solution is to create an FSx for Windows File Server file system in the DR Region and establish connectivity between the VPCs in both Regions by using VPC peering. This will ensure that the data does not travel across the public internet and meets the compliance requirements. By using AWS DataSync with interface VPC endpoints and AWS PrivateLink, the data can be copied securely and efficiently between the FSx for Windows File Server file systems in both Regions. This solution also provides the ability to fail over to the DR Region in case of a disaster. References: [Amazon FSx for Windows File Server User Guide], [AWS DataSync User Guide] , [Amazon VPC User Guide]

https://repost.aws/knowledge-center/ecs-fargate-service-auto-scaling

" The S3 buckets have millions of objects " If there are million of objects then you should use Batch operations. https://aws.amazon.com/blogs/storage/encrypting-objects-with-amazon-s3-batch-operations/

B is correct because SageMaker AI multi-model endpoints are designed for hosting a large number of models that use the same ML framework behind shared serving infrastructure. This is exactly the pattern for a SaaS platform with thousands of tenant-specific models. A dedicated endpoint per model would create excessive cost and management overhead. Serverless endpoints reduce instance management but do not solve the cost problem of managing thousands of separate model deployments. Asynchronous inference and batch transform are not appropriate because the requirement includes sub-200 ms latency, which requires real-time inference. AWS documentation states that multi-model endpoints are ideal for hosting many models using the same framework and can provide higher cost savings by sharing resources across models.

A. Modify the application deployment by building a Docker image that contains the application code. Publish the image to Amazon Elastic Container Registry (Amazon ECR). - This step is necessary to package the application code in a container and make it available for running on ECS. B. Create a new Amazon Elastic Container Service (Amazon ECS) task definition with a compatibility type of AWS Fargate. Configure the task definition to use the new image in Amazon Elastic Container Registry (Amazon ECR). Adjust the Lambda function to invoke an ECS task by using the ECS taskdefinition when a new file arrives in Amazon S3.

Creating a quota request template in the us-east-1 Region of the management account ensures it applies to all newly provisioned accounts in the organization.

By specifying the required ml.p2.xlarge training job and endpoint usage quotas in ap-southeast-2, the accounts will automatically receive these higher quotas upon creation, supporting the student lab environments without additional manual quota requests.

This process integrates seamlessly with AWS Organizations for automated and standardized account provisioning.

Amazon Connect is a cloud-based contact center service that allows you to set up a virtual call center for your business. It provides an easy-to-use interface for managing customer interactions through voice and chat. Amazon Connect integrates with other AWS services, such as Amazon S3 and Amazon Kinesis, to help you collect, store, and analyze customer data for insights into customer behavior and trends. On the other hand, Amazon Pinpoint is a marketing automation and analytics service that allows you to engage with your customers across different channels, such as email, SMS, push notifications, and voice. It helps you create personalized campaigns based on userbehavior and enables you to track user engagement and retention. While both services allow you to communicate with your customers, they serve different purposes. Amazon Connect is focused on customer support and service, while Amazon Pinpoint is focused on marketing and engagement.

Comprehensive and Detailed Explanation:

Why C is the best answer

This question requires all of the following:

Bidirectional access between two separate AWS Organizations

Access to S3 buckets across accounts and organizational boundaries

Encryption at rest

Logging of all S3 bucket access

Buckets exist in different Regions

Option C is the best match because S3 Access Points are purpose-built to simplify and scale access management for shared S3 datasets. For cross-account or multi-team sharing scenarios, S3 Access Points let you define dedicated access paths and policies instead of overloading a single bucket policy. For encryption at rest, customer managed AWS KMS keys are the strongest fit when cross-account access must be explicitly controlled. For logging bucket-level object access, AWS CloudTrail is the correct service because S3 data events record object-level API activity such as GetObject and PutObject.

Why the other options are not correct

A. S3 Access Grants is not the best fit here

S3 Access Grants is designed primarily to grant S3 access to identities mapped from identity sources and simplify data access at scale. It is not the standard primary design choice for bidirectional bucket sharing between two separate organizations. The exam-focused architectural pattern for controlled cross-account S3 sharing is S3 Access Points plus policies, combined with KMS permissions and CloudTrail data events. Even though SSE-KMS and CloudTrail data events are good parts of the design, the access mechanism in A is not the best answer.

B. VPC networking services do not solve the main requirement

VPC peering, PrivateLink, and endpoint access controls are network connectivity controls. Amazon S3 is not shared between organizations by building VPC peering between their VPCs. Also, VPC Flow Logs record network flow metadata, not S3 object access. That means B fails the logging requirement and uses the wrong control plane for the access-sharing requirement.

D. Replication and SCPs do not provide bidirectional shared access

S3 Cross-Region Replication copies objects; it does not establish bidirectional shared access for both organizations to each other’s buckets. Service Control Policies also do not work the way this option describes. SCPs are guardrails within an AWS Organization; they are not shared between separate organizations to grant object access. In addition, S3 server access logging is not the best answer when the requirement is to collect logs of all S3 bucket access in an auditable way for object-level API actions. CloudTrail data events are the stronger and more exam-aligned choice for logging object access activity.

Key technical reasoning

Cross-account / cross-organization S3 access:

S3 Access Points are commonly used to provide distinct permissions and simplified policy management for different consumers of the same bucket. This is especially useful in multi-account environments.

Encryption at rest:

When access spans accounts or organizations, customer managed KMS keys are preferred because key policies and grants can explicitly authorize external principals. This gives finer control than SSE-S3 when cross-account governance matters.

Logging:

For S3 object-level operations, CloudTrail data events are the standard answer in AWS architecture questions. They record API activity such as GetObject, PutObject, and DeleteObject. This is more appropriate than VPC Flow Logs or relying only on S3 server access logging.

Regional aspect:

The fact that buckets are in us-east-1 and us-west-1 does not prevent cross-account access. S3 access control, KMS permissions, and CloudTrail can be configured in the relevant Regions.

The company is building an active-active, multi-Region architecture for an ECS-based web application behind ALBs, with shared container images in ECR and a relational database in Aurora MySQL. The goal is to duplicate the setup in a second Region with minimal operational overhead while allowing minor replication latency.

A is required because ACM certificates are Regional for use with Regional resources like an Application Load Balancer. An ALB in the secondary Region needs an ACM certificate that exists in that same Region to terminate HTTPS. Requesting (or importing) a certificate in the secondary Region and attaching it to the ALB HTTPS listener is the standard managed approach with low operational overhead.

B is not correct because ACM certificates are not shared “across Regions” for direct use by a Regional ALB in another Region. Sharing via AWS RAM is used for sharing resources across accounts within a Region (and for supported resource types), but it does not provide a mechanism to use an ACM certificate issued in one Region directly on an ALB in a different Region.

D is required because the application runs in both Regions and must be able to pull container images locally with minimal operational work. Amazon ECR supports cross-Region replication so images pushed in the primary Region can be replicated automatically to the secondary Region’s ECR repositories. This avoids building custom copy pipelines and reduces the chance of drift. Once replication is enabled, updating images in the primary Region can automatically populate the secondary Region repository.

C is not the best fit. A VPC endpoint for ECR is about private connectivity to ECR within a Region. It does not solve the multi-Region image availability problem by itself, and it introduces the idea of pulling images from the primary Region, which is unnecessary and less optimal when cross-Region replication can keep images local to each Region.

E is required because the database must support multi-Region application operation with tolerable replication latency. Aurora Global Database is the managed feature designed to replicate an Aurora cluster to a secondary Region with low-latency replication and minimal operational overhead compared to custom replication. Enabling write forwarding allows applications in the secondary Region to send write requests to the secondary cluster endpoint, with Aurora forwarding those writes to the primary writer. This reduces application changes compared to forcing all writes to go directly to the primary Region endpoints, and it supports an active-active application deployment pattern where both Regions can accept traffic.

F is not correct because “cross-Region writes between DB clusters” and “multiple writer instances in the primary Region” do not provide a managed, multi-Region, multi-writer Aurora design as described. Aurora’s managed multi-Region pattern is Aurora Global Database with a single primary writer in one Region (with the option to promote a secondary Region during failover) and optional write forwarding from a secondary Region to the primary. The option’s described cross-Region multi-writer behavior is not a standard supported Aurora configuration and would not be the least operational overhead approach.

Therefore, the least operational overhead combination that correctly addresses HTTPS termination in the secondary Region, image availability in both Regions, and managed multi-Region database replication is A, D, and E.

https://docs.aws.amazon.com/global-accelerator/latest/dg/about-accelerators.alb-accelerator.html Option A is wrong. AWS WAF does not support associating with NLB. https://docs.aws.amazon.com/waf/latest/developerguide/waf-chapter.html Option B is wrong. An ALB does not support an Elastic IP address.https://aws.amazon.com/elasticloadbalancing/features/

https://docs.aws.amazon.com/autoscaling/ec2/userguide/adding-lifecycle-hooks.html

- Refer to Default Result section - If the instance is terminating, both abandon and continue allow the instance to terminate. However, abandon stops any remaining actions, such as other lifecycle hooks, and continue allows any other lifecycle hooks to complete.

https://aws.amazon.com/blogs/infrastructure-and-automation/run-code-before-terminating-an-ec2-auto-scaling-instance/

https://github.com/aws-samples/aws-lambda-lifecycle-hooks-function

https://github.com/aws-samples/aws-lambda-lifecycle-hooks-function/blob/master/cloudformation/template.yaml

To develop a migration plan with accurate inventory and dependency data:

AWS Migration Hub provides a single view for tracking migration tasks and resources across multiple AWS services.

The AWS Application Discovery Agent (installed on servers) collects detailed data about running processes, system performance, and network connections.

Migration Hub Strategy Recommendations leverages this data to automatically identify application patterns, generate recommended AWS target services, and provide a detailed migration plan.This approach ensures accurate data collection, detailed dependency mapping, and tailored recommendations—crucial for a successful and right-sized migration to AWS.

C is correct because the root cause is synchronous chaining of Lambda functions that causes the initial function to wait and time out. AWS Step Functions is designed to orchestrate multiple steps (including Lambda invocations) with managed state, retries, and timeouts at the workflow level. By moving the chaining logic into a Step Functions state machine and removing direct function-to-function invocations, the initial Lambda no longer needs to remain running while waiting for downstream functions. Step Functions can coordinate the sequence (or parallel branches), handle transient failures with retries/backoff, and maintain execution state without requiring a long-lived “waiting” Lambda invocation. This resolves timeouts while preserving the operational and cost advantages of a serverless Lambda-based architecture and typically requires minimal code change compared to rewriting or migrating platforms.

Why the other options are incorrect:

A: Migrating to containers/EKS is a major architectural change and increases operational overhead, violating “minimum application changes” and the desire to keep Lambda benefits.

B: Rewriting in Java is extensive and unnecessary. Adding SQS between functions changes invocation style and error handling and still requires significant redesign (and does not inherently solve orchestration/branching as cleanly as Step Functions).

D: Grouping functions into containers and running on ECS Spot changes the compute model and introduces more operational overhead and larger application changes than adopting Step Functions.

CloudFront origin failover is implemented by using an origin group with a primary origin and a secondary origin. CloudFront can automatically fail over to the secondary origin when the primary origin fails under configured conditions. The default origin connection behavior can take up to 30 seconds because CloudFront uses multiple connection attempts and timeout values. AWS documentation states that reducing the connection timeout and reducing the number of connection attempts can make failover occur faster. Option C correctly adds the replicated S3 bucket as a second origin, creates an origin group, updates the default cache behavior to use that origin group, and sets both connection timeout and connection attempts to 1. Option B misses updating the cache behavior and uses the wrong timeout. Lambda-based promotion is unnecessary operational overhead.

This option allows the solutions architect to use Aurora global database to replicate data across multiple AWS Regions with low latency and high availability1. By launching the solution in the target Region, the solutions architect can ensure that the API Gateway, Lambda functions, and other resources are ready to serve traffic in case of a disaster in the source Region. By configuring the two Regional solutions to work in an active-passive configuration, the solutions architect can minimize costs and avoid data conflicts by having only one primary Region that accepts write operations and one secondary Region that serves as a standby2. The RTO and RPO requirements can be met by using Aurora global database, which supports sub-second failover times and near real-time replication1.

Working with Amazon Aurora global database

Active-passive failover

A. High performance computing (HPC) workload cluster should be in a single AZ.

C. Elastic Fabric Adapter (EFA) is a network device that you can attach to your Amazon EC2 instances to accelerate High Performance Computing (HPC)

F. Amazon FSx for Lustre - Use it for workloads where speed matters, such as machine learning, high performance computing (HPC), video processing, and financial modeling.

Cluster – packs instances close together inside an Availability Zone. This strategy enables workloads to achieve the low-latency network performance necessary for tightly-coupled node-to-node communication that is typical of HPC applications.

https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/placement-groups.html

Generally, unstructured data should be converted structured data before querying them. AWS Glue can do that. https://docs.aws.amazon.com/glue/latest/dg/schema-relationalize.html https://docs.aws.amazon.com/athena/latest/ug/glue-athena.html

Using a DHCP options set in AWS allows EC2 instances to automatically receive network configuration information, including the IP address of the on-premises NTP server.

This ensures that all EC2 instances in the VPC consistently use the on-premises atomic clock as their authoritative time source, maintaining the same time synchronization as the existing environment.

By allowing NTP traffic over the Direct Connect connection and ensuring proper security group and network ACL configurations, this approach eliminates the need to manually configure each EC2 instance’s NTP settings, significantly reducing administrative overhead and ensuring consistent, accurate time across the environment.

This aligns with AWS best practices for extending on-premises time synchronization to EC2 instances in a hybrid environment.

Setting up an AWS Client VPN endpoint and integrating it with Active Directory Domain Services (AD DS) using an AD Connector directory enables secure remote access to internal services deployed in a VPC. Enabling multi-factor authentication (MFA) for AD Connector enhances security by adding an additional layer of authentication. This solution meets the company ' s requirements for secure remote access through a VPN with MFA, ensuring that the security policy is adhered to while providing a seamless experience for the remote engineers.

AWS Documentation on AWS Client VPN and AD Connector provides detailed instructions on setting up a Client VPN endpoint and integrating it with existing Active Directory for authentication. This solution aligns with AWS best practices for secure remote access to AWS resources.

The cleanest design is an AWS Config organization conformance pack deployed centrally. Conformance packs are YAML templates that can include AWS Config managed rules, custom rules, and remediation actions. AWS Config also supports managing conformance packs across all member accounts in an AWS Organizations organization, allowing centralized deployment, update, and deletion. The PutOrganizationConformancePack API deploys a conformance pack across member accounts and can create AWS Config rules and remediation actions without requiring member-account IAM permissions for those individual rule operations. Because the requirement includes four custom rules and five managed rules, the conformance pack must include all nine rules. Deploying only custom rules leaves the managed rules outside the update model. CloudFormation StackSets add unnecessary operational overhead and do not provide the same native organization conformance-pack governance.

A is correct because AWS recommends using oneconnection alias per Region, associated witheach directory. Then, configure a Route 53 failover policy so that if the primary Region becomes unhealthy, users are directed to the failover Region automatically. “Evaluate Target Health” ensures automatic detection and failover.

This solution requires minimal operational overhead, as it only requires setting up AWS IoT Core and creating a thing for each device. (Reference: AWS Certified Solutions Architect - Professional Official Amazon Text Book, Page 537)

AWS IoT Core is a fully managed service that enables secure, bi-directional communication between internet-connected devices and the AWS Cloud. It supports the MQTT protocol and includes built-in device authentication and access control. By using AWS IoT Core, the company can easily provision and manage the X.509 certificates for each device, and connect the devices to the service with minimal operational overhead.

Creating a user pool in Amazon Cognito and configuring it for the application will meet the requirement of giving only a specific list of users the ability to access the application from the internet. A user pool is a directory of users that can sign in to an application with ausername and password1. The company can populate the user pool with the required users and configure the pool to require MFA for additional security2. Configuring a listenerrule on the ALB to require authentication through the Amazon Cognito hosted UI will meet the requirement of not changing the application and not integrating it with an identity provider. The ALB can use Amazon Cognito as an authentication action to authenticate usersbefore forwarding requests to the Fargate service3. The Amazon Cognito hosted UI is a customizable web page that provides sign-in and sign-up functionality for users4.

Comprehensive and Detailed Explanation:

This question is mainly about cost optimization for two different workload patterns:

A constant-rate ingestion workload

A noncritical nightly batch analytics workload that can tolerate interruption or rerun

The most cost-effective architecture is to decouple storage from the ingestion servers, store the incoming data in Amazon S3, and run the nightly calculations by using AWS Batch on Spot Instances.

Why B is correct

Amazon Data Firehose can continuously ingest streaming data and deliver the data to Amazon S3. Because the data rate is constant, the company does not need to keep a fleet of EC2 instances running just to collect and retain data on attached EBS volumes. Using S3 as durable storage is operationally simpler and generally more cost-effective than maintaining long-running EC2 ingestion servers with EBS-based storage for this type of streaming landing zone.

For the nightly processing, the job is:

batch oriented,

takes only 4 hours,

not critical to the business,

acceptable to rerun later if interrupted.

These characteristics make Spot Instances an ideal fit. AWS Batch is specifically designed to run batch computing workloads and can efficiently provision and manage compute resources, including Spot Instances. Since the job can tolerate interruption, Spot capacity provides the lowest-cost compute option.

This option reduces cost in both major areas:

storage and ingestion are moved to managed services and object storage,

nightly compute moves from On-Demand Instances to Spot-based batch processing.

Why A is incorrect

A improves the design by moving ingestion to Data Firehose and storage to S3, which is a good step. However, it still uses EC2 On-Demand Instances for nightly processing. Because the nightly workload is not business-critical and can tolerate failure, Spot Instances are a better fit and more cost-effective than On-Demand Instances. Therefore, A is not the most cost-effective choice.

Why C is incorrect

This option keeps ingestion on a fleet of EC2 Reserved Instances with 3-year reservations behind a Network Load Balancer. That is not the most economical or simplest design for constant ingestion when a managed ingestion service such as Data Firehose can deliver directly to S3.

Also, a Network Load Balancer does not add meaningful value for the ingestion requirement described here. The issue is not distributing interactive network traffic to an application fleet. The better architecture is to use a managed data ingestion service instead of recommitting to long-term EC2 reservations.

While AWS Batch with Spot Instances is good for the nightly analytics portion, the ingestion side of this option is still more expensive and less efficient than B.

Why D is incorrect

Amazon Redshift is a data warehouse and is useful for analytical querying, but this option is not the most cost-effective for the stated use case. The company needs to ingest constant-rate streaming market data and run a nightly aggregate job. Storing all incoming data in Redshift is generally a more specialized and potentially more expensive architecture than landing the raw stream in S3 and running elastic batch processing on demand.

The Lambda part is also problematic. The batch processing currently takes 4 hours. AWS Lambda is not designed for long-running heavy batch jobs of this kind. Even if the aggregate query could be expressed in SQL, the answer is still not the most cost-effective and does not align as well with the workload pattern as S3 plus AWS Batch on Spot Instances.

Key design principles being tested

This question tests whether you can identify:

S3 as low-cost, durable storage for incoming streamed data

Managed ingestion services instead of self-managed EC2 ingestion servers

Spot Instances for interruptible, fault-tolerant batch jobs

AWS Batch as the managed orchestration service for batch workloads

Final justification

The most cost-effective redesign is to:

ingest the streaming data with Amazon Data Firehose,

store the data in Amazon S3,

run nightly analytics through AWS Batch on Spot Instances.

That makes B the best answer.

Option B is correct because AWS Organizations allows a company to create a new organization from a chosen payer account and define an organizational unit hierarchy. This way, the finance department can have a centralized method for payment but also maintain visibility into each group’s spending to allocate costs. The company can also invite the existing accounts to join the organization and create new accounts using Organizations, which simplifies the account management process.

Option D is correct because enabling all features of AWS Organizations and establishing appropriate service control policies (SCPs) that filter IAM permissions for sub-accounts allows the security team to have a centralized mechanism to control IAM usage in all the company’s accounts. SCPs are policies that specify the maximum permissions for an organization or organizational unit (OU), and they can be used to restrict access to certain services or actions across all accounts in an organization.

Option A is incorrect because using a collection of parameterized AWS CloudFormation templates defining common IAM permissions that are launched into each account requires more effort than using SCPs. Moreover, it does not provide a centralized mechanism to control IAM usage, as each account would have to launch the appropriate stacks to enforce the least privilege model.

Option C is incorrect because requiring each business unit to use its own AWS accounts does not provide a centralized method for payment or a centralized mechanism to control IAM usage. Tagging each AWS account appropriately and enabling Cost Explorer to administer chargebacks may help with cost allocation, but it is not as efficient as using AWS Organizations.

Option E is incorrect because consolidating all of the company’s AWS accounts into a single AWS account does not provide visibility into each group’s spending or a way to control IAM usage for different business units. Using tags for billing purposes and the IAM’s Access Advisor feature to enforce the least privilege model may help with cost optimization and security, but it is not as scalable or flexible as using AWS Organizations.

AWS Organizations

Service Control Policies

AWS CloudFormation

Cost Explorer

IAM Access Advisor

The best answer is C because the workload has millions of small 64 KB uploads every hour. Sending every S3 object event through EventBridge to an ECS task creates high event and invocation overhead. Amazon Data Firehose can buffer many incoming records and deliver larger compressed objects to S3, which reduces the small-object problem and improves cost efficiency. The ECS processing takes an average of 15 minutes, so replacing it with Lambda is risky because Lambda has a maximum runtime of 900 seconds, exactly 15 minutes. For five-year retention, S3 Lifecycle rules can transition data to S3 Glacier Deep Archive for long-term low-cost archival. DynamoDB TTL is appropriate for automatically expiring 15-day results without custom cleanup jobs. (AWS Documentation)

===============

https://docs.aws.amazon.com/sns/latest/dg/sns-cross-region-delivery.html

Auto Scaling Group with Application Load Balancer:

Moving the Tomcat server to an Auto Scaling group ensures that the number of instances adjusts dynamically based on the traffic load. An Application Load Balancer (ALB) distributes incoming traffic across multiple instances, improving the application ' s reliability and availability.

Migrate to Amazon Aurora with Replica:

Migrating the MySQL database to Amazon Aurora and adding an Aurora Replica enhances the database ' s scalability and availability. Aurora is optimized for performance, and replicas help distribute read traffic, reducing the load on the primary instance.

Additional Public Subnet:

Creating an additional public subnet in a different Availability Zone enhances fault tolerance. This ensures that the website remains accessible even if one Availability Zone experiences issues.

References

AWS Well-Architected Framework

Amazon Aurora Documentation

This solution will meet the requirements of making the application highly available with the least development time. Creating a new AMI that is configured with SSM Agent will enable the company to use AWS Systems Manager to manage and patch the EC2 instances automatically, reducing downtime and human errors. Using a launch template for an Auto Scaling group will allow the company to launch multiple instances of the same configuration and scale them up or down based on demand. Using smaller instances in the Auto Scaling group will reduce the cost and improve the performance of the application tier. Creating an Application Load Balancer to distribute traffic across the instances in the Auto Scaling group will increase the availability and fault tolerance of the application tier. Migrating the database to Amazon Aurora MySQL will provide a fully managed, compatible, and scalable relational database service that can handle high throughput and concurrent connections.

The requirement is to test IPv6 connectivity in an AWS VPC. When a VPC is associated with an IPv6 CIDR block, subnets can be configured to assign IPv6 addresses to resources, and routing must be correctly configured to allow IPv6 traffic to flow.

For public IPv6 connectivity, IPv6 traffic must be routed through an internet gateway. Unlike IPv4, IPv6 addresses are globally routable by default, and there is no concept of NAT for IPv6 egress through a NAT gateway. Therefore, for an EC2 instance in a public subnet that needs inbound and outbound IPv6 connectivity, the subnet route table must include a route that sends IPv6 traffic (::/0) to an internet gateway. Option C correctly describes this configuration and is a standard pattern for enabling IPv6 access for public subnets.

For private subnets that require outbound-only IPv6 connectivity (for example, instances that must initiate connections to the internet but must not accept inbound connections), AWS provides an egress-only internet gateway. An egress-only internet gateway allows outbound IPv6 traffic while blocking unsolicited inbound IPv6 traffic, similar in intent to how a NAT gateway is used for IPv4. Option E correctly uses an egress-only internet gateway for IPv6 traffic from a private subnet, making it the correct choice for private IPv6 connectivity testing.

Option A focuses on Direct Connect integration. Although Direct Connect can support IPv6, associating a virtual private gateway with a Direct Connect gateway is not required simply to test IPv6 connectivity to customers worldwide. This option also does not explicitly configure IPv6 internet routing and therefore does not directly meet the stated requirement.

Option B is incorrect because NAT gateways do not support IPv6. NAT gateways are IPv4-only services, so routing IPv6 traffic to a NAT gateway is not a valid configuration.

Option D is also incorrect because NAT instances do not provide a supported or recommended solution for IPv6 traffic. NAT-based designs are intended for IPv4 address translation and are not used for IPv6 connectivity in AWS.

Therefore, using an internet gateway for public IPv6 subnets and an egress-only internet gateway for private IPv6 subnets is the correct and supported approach.

https://docs.aws.amazon.com/IAM/latest/UserGuide/confused-deputy.html

This is the most secure way to allow org1 to access resources in org2 because it allows for least privilege security access. The customer should create an IAM role and assign the required permissions to the IAM role. The partner company should then use the IAM role’s Amazon Resource Name (ARN) and include the external ID in the IAM role’s trust policy when requesting access to perform the required tasks. This ensures that the partner company can only access the resources that it needs and only from the specific customer account.

The best migration strategy to meet the requirement of minimizing database service interruption before the DNS cutover is to use AWS DMS to migrate data between the two Aurora DB clusters. AWS DMS can perform continuous replication of data with high availability and consolidate databases into a petabyte-scale data warehouse by streaming data to Amazon Redshift and Amazon S31. AWS DMS supports homogeneous migrations such as migrating from one Aurora MySQL DB cluster to another, as well as heterogeneous migrations between different database platforms2. AWS DMS also supports cross-account migrations, as long as the source and target databases are in the same AWS Region3.

The other options are not optimal for the following reasons:

Option A: Taking a snapshot of the existing Aurora database and restoring it in the new account would require a downtime during the snapshot and restore process, which could be significant for large databases. Moreover, any changes made to the source database after the snapshot would not be replicated to the target database, resulting in data inconsistency4.

Option C: Using AWS Backup to share an Aurora database backup from the existing AWS account to the new AWS account would have the same drawbacks as option A, as AWS Backup uses snapshots to create backups of Aurora databases.

Option D: Using AWS Application Migration Service to migrate data between the two Aurora DB clusters is not a valid option, as AWS Application Migration Service is designed to migrate applications, not databases, to AWS. AWS Application Migration Service can migrate applications from on-premises or other cloud environments to AWS, using agentless or agent-based methods.

https://aws.amazon.com/about-aws/whats-new/2022/02/amazon-cloudfront-managed-prefix-list/

The company should deploy a new Amazon Elastic File System (Amazon EFS) Multi-AZ file system. The company should configure the file system for 75 MiBps of provisioned throughput. The company should implement replication to a file system in the DR Region. This solution will meet the requirements because Amazon EFS is a serverless, fully elastic file storage service that lets you share file data without provisioning or managing storage capacity and performance. Amazon EFS is built to scale on demand to petabytes without disrupting applications, growing and shrinking automatically as you add and remove files1. By deploying a new Amazon EFS Multi-AZ file system, the company can create a single location for updates to application data for all instances. A Multi-AZ file system replicates data across multiple Availability Zones (AZs) within a Region, providing high availability and durability2. By configuring the file system for 75 MiBps of provisioned throughput, the company can ensure that it meets the peak operations requirement of 225 MiBps of read throughput. Provisioned throughput is a feature that enables you to specify a level of throughput that the file system can drive independent of the file system’s size or burst credit balance3. By implementing replication to a file system in the DR Region, the company can make a copy of the data available in another AWS Region for disaster recovery. Replication is a feature that enables you to replicate data from one EFS file system to another EFS file system across AWS Regions. The replication process has an RPO of less than 1 hour.

The other options are not correct because:

Deploying a new Amazon FSx for Lustre file system would not provide a single location for updates to application data for all instances. Amazon FSx for Lustre is a fully managed service that provides cost-effective, high-performance storage for compute workloads. However, it does not support concurrent write access from multiple instances. Using AWS Backup to back up the file system to the DR Region would not provide real-time replication of data. AWS Backup is a service that enables you to centralize and automate data protection across AWS services. However, it does not support continuous data replication or cross-Region disaster recovery.

Deploying a General Purpose SSD (gp3) Amazon Elastic Block Store (Amazon EBS) volume with 225 MiBps of throughput would not provide a single location for updates to application data for all instances. Amazon EBS is a service that provides persistent block storage volumes for use with Amazon EC2 instances. However, it does not support concurrent access from multiple instances, unless Multi-Attach is enabled. Enabling Multi-Attach for the EBS volume would not provide Multi-AZ resilience or cross-Region replication. Multi-Attach is a feature that enables you to attach an EBS volume to multiple EC2 instances within the same Availability Zone. Using AWS Elastic Disaster Recovery to replicate the EBS volume to the DR Region would not provide real-time replication of data. AWS Elastic Disaster Recovery (AWS DRS) is a service that enables you to orchestrate and automate disaster recovery workflows across AWS Regions. However, it does not support continuous data replication or sub-hour RPOs.

Deploying an Amazon FSx for OpenZFS file system in both the production Region and the DR Region would not be as simple or cost-effective as using Amazon EFS. Amazon FSx for OpenZFS is a fully managed service that provides high-performance storage with strong data consistency and advanced data management features for Linux workloads. However, it requires more configuration and management than Amazon EFS, which is serverless and fully elastic. Creating an AWS DataSync scheduled task to replicate the data from the production file system to the DR file system every 10 minutes would not provide real-time replication of data. AWS DataSync is a service that enables you to transfer data between on-premises storage and AWS services, or between AWS services. However, it does not support continuous data replication or sub-minute RPOs.

This solution will allow the detection logic to be run as soon as the image is uploaded to the S3 bucket, before it is served to users via the CloudFront distribution. This way, the detection logic can quickly identify any corrupted images and prevent them from being served to users, minimizing latency between ingestion and serving.

The requirement is to reduce application latency for a globally distributed user base, with users primarily in the United States but increasing usage in Europe and the Middle East. The application is read-heavy and uses Amazon DynamoDB as its data store. Reducing latency for global users requires placing both the application compute and the data closer to users in each Region.

Option D uses the correct multi-Region architecture. Deploying the application to multiple AWS Regions places Amazon EKS workloads closer to end users. Converting the DynamoDB table to a global table creates fully managed, multi-Region replicas of the data with low-latency replication. This allows reads to be served locally from the nearest DynamoDB replica, significantly reducing read latency for users outside the primary Region. Because the workload is read-heavy, this design is particularly effective.

Route 53 geolocation routing directs users to the Region closest to them based on their geographic location. Combined with health checks and Evaluate Target Health, this ensures users are routed to the nearest healthy Region, improving both latency and availability.

Option A is incorrect because global secondary indexes do not replicate data across Regions. GSIs operate within a single Region and are used to support alternative query patterns, not global data distribution. Additionally, Route 53 failover routing is designed for disaster recovery, not for performance optimization or global load balancing.

Option B uses DynamoDB Accelerator (DAX), which improves read latency by caching data in memory, but DAX is Region-specific and does not reduce cross-Region network latency for users in Europe and the Middle East accessing a US-based application. It also does not address application-level latency caused by distance to the EKS cluster.

Option C is not valid because DynamoDB does not support read replicas in the same way relational databases do. DynamoDB replication for latency and global access is achieved through global tables, not through read replicas in a single Region. CloudFront can reduce latency for static or cacheable HTTP content, but it does not solve backend application latency or DynamoDB access latency for dynamic, read-heavy requests.

Therefore, deploying the application and DynamoDB as global tables across Regions and using Route 53 geolocation routing is the correct solution to minimize latency for all global users.

https://docs.aws.amazon.com/IAM/latest/UserGuide/reference_policies_examples_aws_deny-requested-region.html

https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps_examples_ec2.html

https://docs.aws.amazon.com/AmazonS3/latest/userguide/storage_lens.html

Enable DynamoDB Auto Scaling:

Navigate to the DynamoDB console and select the table experiencing high demand.

Go to the " Capacity " tab and enable auto scaling for both read and write capacity units. Auto scaling adjusts the provisioned throughput capacity automatically in response to actual traffic patterns, ensuring the table can handle the increased load.

Configure Auto Scaling Policies:

Set the minimum and maximum capacity units to define the range within which auto scaling can adjust the provisioned throughput.

Specify target utilization percentages for read and write operations, typically around 70%, to maintain a balance between performance and cost.

Monitor and Adjust:

Use Amazon CloudWatch to monitor the auto scaling activity and ensure it is effectively handling the increased demand.

Adjust the auto scaling settings if necessary to better match the traffic patterns and application requirements.

By enabling DynamoDB auto scaling, you ensure that the database can handle the fluctuating traffic volumes without manual intervention, improving response times and reducing errors.

References

AWS Compute Blog on Using API Gateway as a Proxy for DynamoDB【60】.

AWS Database Blog on DynamoDB Accelerator (DAX)【59】.

The company should create a Network Load Balancer (NLB) and associate it with one static IP address in multiple Availability Zones. The company should also create an ALB-type target group for the NLB and add the existing ALB. The company should add the NLB IP addresses to the firewall appliance and update the clients to connect to the NLB. This solution will allow traffic flow to AWS from the on-premises network by using static IP addresses that can be added to the firewall appliance’s allow list. The NLB will forward requests to the ALB, which will use path-based routing to forward requests to the target groups.

For an application requiring low-latency access to financial information and high availability with a Recovery Time Objective (RTO) of 30 seconds, using Amazon DynamoDB for data storage and Amazon QuickSight for reporting is the most suitable solution. DynamoDB offers fast, consistent, and single-digit millisecond latency for data retrieval, meeting the latency requirements. QuickSight ' s ability to directly query DynamoDB datasets and provide embedded dashboards for reporting enables real-time financial report generation. This combination ensures high availability and meets the RTO requirement, providing a robust solution for the application ' s needs.

Amazon DynamoDB Documentation: Describes the features and benefits of DynamoDB, emphasizing its performance and scalability for applications requiring low-latency access to data.

Amazon QuickSight Documentation: Provides information on using QuickSight for creating and embedding interactive dashboards, including direct querying of DynamoDB datasets for real-time data visualization.

Option C is correct. For an internet-facing Application Load Balancer used as a CloudFront custom origin, AWS recommends configuring CloudFront to add a custom HTTP header to origin requests and configuring the ALB listener rules to forward only requests that contain that specific header. This prevents users from bypassing CloudFront and accessing the ALB directly, while still allowing legitimate CloudFront-originated traffic. AWS also warns that the header name and value must remain secret and should be protected with HTTPS. Option A is incomplete because security groups cannot restrict traffic to one specific CloudFront distribution. Option B does not reliably stop direct ALB access. Option D is wrong because CloudFront origin access control is primarily for supported AWS origins such as S3 origins, not this ALB custom-origin control pattern.

This option meets the RPO and RTO requirements for both the application and database tiers and uses tools like Amazon DLM and RDS automated backups to create and manage the backups. Additionally, it uses Global Accelerator to ensure low latency after failover by directing traffic to the closest healthy endpoint.

You can segment your network by creating multiple route tables in an AWS Transit Gateway and associate Amazon VPCs and VPNs to them. This will allow you to create isolated networks inside an AWS Transit Gateway similar to virtual routing and forwarding (VRFs) in traditional networks. The AWS Transit Gateway will have a default route table. The use of multiple route tables is optional.

S3 Transfer Acceleration is a feature that enables fast, easy, and secure transfers of files over long distances between your client and an S3 bucket1. It works by leveraging the CloudFront edge network to route your requests to S3 over an optimized network path1. By using a Transfer Acceleration endpoint when generating a presigned URL, you can allow authenticated users to upload objects fasterand more reliably2. Additionally, using the S3 multipart upload API can improve the performance of large object uploads by breaking them into smaller parts and uploading them in parallel3.

S3 Transfer Acceleration

Using Transfer Acceleration with presigned URLs

Uploading objects using multipart upload API

Amazon Elastic Container Service (ECS) on AWS Fargate is a fully managed service that allows you to run containers without having to manage the underlying infrastructure. When you launch tasks on Fargate, resources are automatically allocated based on the number of tasks running, which reduces the operational overhead.

Using ECS on Fargate allows you to assign custom tags to tasks using the --tags option in the run-task command, as described in the documentation:https://docs.aws.amazon.com/cli/latest/reference/ecs/run-task.htmlYou can also set enableECSManagedTags to true, which allows the service to automatically add the cluster name and service name as tags.https://docs.aws.amazon.com/AmazonECS/latest/developerguide/task-placement-constraints.html#tag-based-scheduling

This option uses the S3 Storage Lens default dashboard to track bucket and encryption metrics across two AWS Regions. S3 Storage Lens is a feature that provides organization-wide visibility into object storage usage and activity trends, and delivers actionable recommendations to improve cost-efficiency and apply data protection best practices. S3 Storage Lens delivers more than 30 storage metrics, including metrics on encryption, replication, and data protection. The default dashboard provides a summary of the entire S3 usage and activity across all Regions and accounts in an organization. The company can give the compliance teams access to the dashboard directly in the S3 console, which requires the least operational overhead.

Cross-zone load balancing enables traffic to be distributed evenly across all registered instances in all enabled Availability Zones. However, this also increases data transfer charges between Availability Zones. By turning off cross-zone load balancing, the serviceprovider applications can reduce inter-Availability Zone data transfer costs. Similarly, by using the Availability Zone-specific endpoint service, the service consumer applications can ensure that they connect to the nearest service provider application in the same Availability Zone, avoiding cross-Availability Zone data transfer charges. References:

https://docs.aws.amazon.com/elasticloadbalancing/latest/network/load-balancer-target-groups.html#cross-zone-load-balancing

https://docs.aws.amazon.com/vpc/latest/userguide/vpce-interface.html#vpce-interface-dns

https://docs.aws.amazon.com/systems-manager/latest/userguide/systems-manager-patch.html

Amazon AppStream 2.0 offers a streamlined solution for rehosting a Windows-based desktop application on AWS with minimal development effort. By creating an AppStream 2.0image that includes the application and using an On-Demand fleet for streaming, the application becomes accessible from any device, including Linux machines. AppStream 2.0 user pools can be used for authentication, simplifying access management without the need for extensive changes to the application or infrastructure.

AWS Documentation on Amazon AppStream 2.0 provides insights into setting up application streaming solutions. This approach is recommended for delivering desktop applications to diverse operating systems without the complexity of managing virtual desktops or extensive application refactoring.

This explanation is based on AWS documentation and best practices but is paraphrased, not a literal extract.

The current CI/CD pipeline uses an IAM user with long-lived access keys stored in GitHub Actions. The new requirement is that pipelines must not use long-lived secret keys. Instead, the solution should provide short-lived credentials with minimal operational overhead.

GitHub Actions natively supports integration with cloud providers using OpenID Connect (OIDC). With OIDC, GitHub acts as an identity provider that can issue OIDC tokens to workflows. On the AWS side, IAM supports configuring an OIDC identity provider and roles that can be assumed by principals presenting valid OIDC tokens through the sts:AssumeRoleWithWebIdentity API. This pattern enables short-lived, automatically rotated credentials for CI/CD jobs without storing long-lived secrets.

In the correct solution (option B), you configure an IAM OIDC identity provider for GitHub in the AWS account. You then create a new IAM role with a trust policy that allows the Github OIDC provider to call sts:AssumeRoleWithWebIdentity, with conditions that restrict which repositories or workflows can assume the role. The existing IAM policy that grants deployment permissions is attached to that role. In GitHub Actions, you update the pipeline configuration to request an OIDC token and assume the IAM role at runtime. Each workflow run receives short-lived credentials without storing static keys, and AWS automatically handles the token verification and temporary credential issuance. This approach is the AWS-recommended pattern for integrating GitHub Actions with AWS without long-lived secrets and has low operational overhead once configured.

Option A uses SAML 2.0, which is typically used for enterprise single sign-on for users, not for GitHub Actions workflows. GitHub does not natively use SAML to obtain AWS credentials for CI/CD pipelines in the same streamlined way as OIDC, and implementing a SAML-based integration would add unnecessary complexity.

Option C introduces Amazon Cognito as an indirection layer. Although Cognito can federate with external identity providers, including social providers, using it as an intermediary to obtain temporary AWS credentials for a machine-to-machine CI/CD pipeline is not necessary when IAM OIDC federation with GitHub is directly supported. This adds additional configuration and operational overhead.

Option D uses IAM Roles Anywhere with client certificates from AWS Private CA. Roles Anywhere is designed for workloads running outside AWS that need to assume IAM roles using X.509 certificates instead of access keys. While technically possible, it requires managing private certificates, trust anchors, and a credential helper tool, which is more complex and operationally heavier than the direct OIDC integration specifically designed for GitHub Actions.

Therefore, configuring an IAM OIDC identity provider for GitHub and creating an IAM role to be assumed via sts:AssumeRoleWithWebIdentity (option B) meets the requirement to replace long-lived secret keys with short-lived credentials with the least operational overhead.

Utilizing an IP access control group rule with the list of public addresses from branch offices and associating it with the Amazon WorkSpaces directory is the most operationally efficient solution. This method ensures that access to WorkSpaces is restricted to specified locations, aligning with the corporate security policy. This approach offers simplicity and flexibility, especially with the potential addition of a new branch office, as updating the IP access control group is straightforward.

AWS Documentation on Amazon WorkSpaces and IP Access Control Groups provides detailed instructions on how to implement access restrictions based on IP addresses. This solution aligns with best practices for securing virtual desktops while maintaining operational efficiency.

https://docs.aws.amazon.com/vpc/latest/privatelink/vpc-endpoints-access.html

https://aws.amazon.com/premiumsupport/knowledge-center/vpc-reduce-nat-gateway-transfer-costs/

VPC endpoint policies enable you to control access by either attaching a policy to a VPC endpoint or by using additional fields in a policy that is attached to an IAM user, group, or role to restrict access to only occur via the specified VPC endpoint

The application is business-critical with an SLA requirement of 99.95% uptime and is currently limited by an on-premises data center. The architecture includes a PostgreSQL database and separate business logic and presentation layers. Remote users experience high latency when accessing the application, which is latency-sensitive.

To meet or exceed a 99.95% SLA for the database while minimizing operational burden, a managed database service with built-in high availability is preferred. Amazon Aurora PostgreSQL-compatible editions provide high availability and durability by automatically replicating data across multiple Availability Zones in a Region. Aurora can offer higher availability and faster failover compared to self-managed databases on EC2, and offloads patching, backups, and maintenance.

For the application and presentation layers, running them on Amazon EC2 instances in an Auto Scaling group behind an Application Load Balancer preserves the basic architecture (stateless application servers behind a load balancer) while moving to a managed, highly available environment that can scale out and in automatically based on demand. This results in minimal changes to the application components and provides improved resilience compared to on-premises VMs.

Remote users experience slow load times because their desktop clients connect over higher-latency networks to the on-premises servers. By using Amazon AppStream 2.0, the application can be streamed from AWS to end users. AppStream 2.0 runs the desktop application close to the backend database and application servers within AWS, and streams only the display and input over the network. This significantly reduces the effect of latency between the user and the application backend, improving user experience, especially for latency-sensitive desktop applications, without requiring major changes to the application itself.

Option A uses a PostgreSQL database on EC2, which requires the customer to manage availability, backups, patching, and failover. This does not provide the same managed high-availability guarantees as Aurora or RDS Multi-AZ. It also suggests allocating a full Amazon WorkSpaces desktop per user, which can be more expensive and more complex than streaming only the application via AppStream 2.0.

Option C uses Amazon RDS for PostgreSQL with Multi-AZ configuration, which provides high availability and is a valid managed option. However, it runs the application and presentation layers on Fargate containers behind a Network Load Balancer. NLB is typically used for TCP-level load balancing and is better suited for protocols that require low-level handling, whereas HTTP/HTTPS web applications are commonly placed behind an Application Load Balancer, which provides advanced HTTP routing features. Also, ElastiCache can reduce database load and improve response times but does not directly solve the end-user network latency issue in a desktop client scenario.

Option D is incorrect because Amazon Redshift is a data warehouse service, not an operational PostgreSQL-compatible database for transaction processing. It is not appropriate for hosting the application’s primary relational database. CloudFront accelerates delivery of static and cached web content, not interactive desktop client-server traffic.

Therefore, migrating the database to Aurora PostgreSQL, running the application and presentation layers on EC2 Auto Scaling behind an Application Load Balancer, and using Amazon AppStream 2.0 to deliver the application to remote users (option B) meets the availability requirements, improves user experience, requires relatively little change to the application architecture, and minimizes operational costs compared to more complex or less-managed alternatives.

https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_org_support-all-features.html

https://docs.aws.amazon.com/singlesignon/latest/userguide/get-started-prereqs-considerations.html

Set up an Amazon Aurora MySQL database. Use AWS Database Migration Service (AWS DMS) to perform continuous data replication from the on-premises database to Aurora. Create an Aurora Replica for the Aurora MySQL database, and move the aggregation jobs to run against the Aurora Replica. Set up collection endpoints as AWS Lambda functions behind an Application Load Balancer (ALB), and use Amazon RDS Proxy to write to the Aurora MySQL database. When the databases are synced, point the collector DNS record to the ALB. Disable the AWS DMS sync task after the cutover from on premises to AWS.

Amazon RDS Proxy allows applications to pool and share connections established with the database, improving database efficiency and application scalability. With RDS Proxy, failover times for Aurora and RDS databases are reduced by up to 66%

Option A is correct because AWS IoT Core is a managed MQTT broker service built to handle large-scale device connectivity and message ingestion without the company managing Mosquitto brokers on EC2. AWS IoT rules can process MQTT topic messages and route them to AWS services, including AWS Step Functions. A Step Functions Express workflow is suitable for high-volume, event-driven processing because it is designed for short-duration, high-throughput workflows. Option B is not the best fit because EventBridge Pipes is not the direct AWS IoT rules action pattern required here. Option C requires changing devices to send data to Firehose instead of MQTT. Option D also requires device-side protocol changes and does not provide the managed MQTT connectivity layer that solves the broker scalability problem.

Integrate ALB with OIDC IdP:

In the AWS Management Console, navigate to the Application Load Balancer (ALB) settings.

Configure the ALB to use the OpenID Connect (OIDC) IdP for authentication. This ensures that all requests routed through the ALB are authenticated using the IdP.

Set Up Authentication Rules:

Create a listener rule on the ALB that requires authentication. This rule will forward requests to the IdP for user authentication before allowing access to the backend services.

Restrict Unauthenticated Access:

Ensure the ALB only forwards requests to backend services if the user is authenticated. Unauthenticated requests should be blocked or redirected to the IdP for authentication.

Update CloudFront Configuration:

Modify the CloudFront distribution to forward authenticated requests to the ALB. Ensure that the ALB and API Gateway accept only requests coming through the CloudFront distribution to enforce consistent authentication and security.

By enforcing authentication at the ALB level, you ensure that all backend services are accessed only by authenticated users, enhancing the overall security of the web application

Hosting the .NET Core components on Amazon ECS with the Amazon EC2 launch type will meet the requirements of having complex orchestration and full control over networking and host configuration. Amazon ECS is a fully managed container orchestration service that supports both AWS Fargate and Amazon EC2 as launch types. The Amazon EC2 launch type allows users to choose their own EC2 instances, configure their own networking settings, and access their own host operating systems. Hosting the database on Amazon Aurora MySQL Serverless v2 will meet the requirements of having a strongly relational database model and using the same database engine as SQL Server. MySQL is a compatible relational database engine with SQL Server, and it can support most of the legacy application’s database model. Amazon Aurora MySQL Serverless v2 is a serverless version of Amazon Aurora MySQL that can scale up and down automatically based on demand. Using Amazon SQS for asynchronous messaging will meet the requirements of providing a compatible replacement for MSMQ, which is a queue-based messaging system3. Amazon SQS is a fully managed message queuing service that enables decoupled and scalable microservices, distributed systems, and serverless applications.

The requirement is preventive enforcement during CloudFormation provisioning. AWS CloudFormation Hooks are designed to evaluate CloudFormation and Cloud Control API operations before resources are created or updated. If a resource is noncompliant, a Hook can fail the operation and prevent provisioning. Lambda Hooks allow custom logic, which fits predefined company-specific rules. Deploying and protecting Hooks centrally with StackSets, SCPs, and IAM policies gives consistent enforcement across the organization with low operational overhead. AWS Config custom rules are detective controls; they evaluate resources after deployment and cannot reliably prevent initial creation. CloudFormation macros require templates to explicitly use the macro and are not a universal enforcement layer. Manually invoking CloudFormation Guard through a Lambda function does not automatically block every CloudFormation deployment unless separately integrated.

This option allows the company to use a private repository in Amazon ECR to store and manage its Docker images securely and efficiently1. By creating a permissions policy for the repository that allows only required ECR operations, such as ecr:GetDownloadUrlForLayer, ecr:BatchGetImage, ecr:BatchCheckLayerAvailability, ecr:PutImage, and ecr:InitiateLayerUpload2, the company can restrict access to the repository and prevent unauthorized actions. By including a condition to allow the ECRoperations if the value of the aws:PrincipalOrgID condition key is equal to the ID of the company’s organization, the company can ensure that only accounts that are within its organization can access the images3. By adding a lifecycle rule to the ECR repository that deletes all untagged images over the count of five, the company can reduce storage costs and retain only the most recent untagged images4.

Amazon ECR private repositories

Amazon ECR repository policies

Restricting access to AWS Organizations members

Amazon ECR lifecycle policies

Option B is correct because AWS Client VPN provides managed, secure remote access from individual users to resources inside a VPC without exposing the application directly to the public internet. The application runs in private subnets, so remote employees need authenticated VPN connectivity into the VPC rather than inbound public access. AWS Client VPN endpoints are associated with target networks, which are VPC subnets, and authorization rules determine what connected clients can access. Associating private subnets across Availability Zones also supports availability for client connectivity. A Site-to-Site VPN is designed for network-to-network connectivity, not unmanaged individual remote users. Network ACLs and security group allow lists do not solve the “no direct public internet connection” requirement. AWS documentation confirms that Client VPN endpoints use associated target networks and authorization rules to provide VPC access.

It is the fastest when it comes to rollback and deploying changes every hour

Configure AWS Budgets in the organization ג €™s master account and configure budget alerts that are grouped by application, environment, and owner. Add each business unit to an Amazon SNS topic for each alert. Use Cost Explorer in the organization ג €™s master account to create monthly reports for each business unit.

https://aws.amazon.com/about-aws/whats-new/2019/07/introducing-aws-budgets-reports/#:~:text=AWS%20Budgets%20gives%20you%20the,below%20the%20threshold%20you%20define

This option allows the solutions architect to use a DeletionPolicy attribute to specify how AWS CloudFormation handles the deletion of an S3 bucket when the stack is deleted1. By setting the value of Delete, the solutions architect can instruct CloudFormation to delete the bucket and all of itscontents1. This option does not require any major changes to the application’s architecture or any additional resources.

Deletion policies

By creating an AMI that has Grafana pre-installed and storing the existing dashboards in Amazon Elastic File System (Amazon EFS) it allows for faster and more efficient scaling, and by creating an Auto Scaling group that uses the new AMI and setting the Auto Scaling group ' s minimum, desired, and maximum number of instances to one and creating an Application Load Balancer that serves at least two Availability Zones, it ensures high availability and minimized downtime.

The best solution is to create a tag policy that contains the allowed project tag values in the organization’s management account and create an SCP that denies the cloudformation:CreateStack API operation unless a project tag is added. A tag policy is a type of policy that can help standardize tags across resources in the organization’s accounts. A tag policy can specify the allowed tag keys, values, and case treatment for compliance. A service control policy (SCP) is a type of policy that can restrict the actions that users and roles can perform in the organization’s accounts. An SCP can deny access to specific API operations unless certain conditions are met, such as having a specific tag. By creating a tag policy in the management account and attaching it to each OU, the organization can enforce consistent tagging across all accounts. By creating an SCP that denies the cloudformation:CreateStack API operation unless a project tag is added, the organization can prevent users from creating new resources without proper tagging. This solution will meet the requirements with the least effort, as it does not involve creating additional resources or modifying existing ones. References: Tag policies - AWS Organizations, Service control policies - AWS Organizations, AWS CloudFormation User Guide

B and E are correct. A multi-node Amazon EKS cluster provides high availability for the Kubernetes control and workload layer, while Amazon Managed Service for Prometheus and Amazon Managed Grafana reduce monitoring overhead compared with self-managed Prometheus and Grafana. Amazon ECR enhanced scanning is also appropriate because it uses Amazon Inspector to scan container images for operating system and programming language package vulnerabilities. For storage, Amazon FSx for NetApp ONTAP is the correct multiprotocol shared file system because it supports access patterns such as NFS and SMB through a fully managed service. EFS plus FSx for Windows splits storage across two services and does not provide one multiprotocol storage platform. A single-node EKS cluster is not highly available.

D is correct becauseRDS for PostgreSQLsupportsforeign data wrappers (FDW)that allow querying remote Oracle databases. WithAWS Schema Conversion Tool (SCT)andDatabase Migration Service (DMS), schema and data can be migrated effectively.AWS Direct Connectensures secure, private connectivity to on-prem databases. Cron jobs can be run via EventBridge or external orchestration.

A doesn ' t support relational/spatial querying.

B doesn’t support FDW or spatial types.

C introduces unnecessary complexity.

This solution allows for deploying the Lambda function and API Gateway endpoint to another region, providing a failover option in case of any issues in the primary region. Using Route 53 ' s failover routing policy allows for automatic routing of traffic to the healthy endpoint, ensuring that the application is available even in case of issues in one region. This solution provides a cost-effective and simple way to implement failover while minimizing operational overhead.

https://aws.amazon.com/premiumsupport/knowledge-center/aws-batch-requests-error/

https://aws.amazon.com/premiumsupport/knowledge-center/api-gateway-429-limit/

The main problem is that one client is generating a large number of PUT requests, resulting in increased errors. Because the clients are authenticated by API keys and the problematic traffic is concentrated among a small set of clients, API Gateway usage plans and throttling are a direct control mechanism to limit the request rate per client (or per API key). The API is noncritical and clients can tolerate retries, so returning throttling responses is acceptable as long as clients handle them correctly. The goal is to reduce visible customer-facing errors and protect the API from being overwhelmed by a single client.

API Gateway usage plans can set request throttling limits (rate and burst) per API key. When a client exceeds the configured throttle, API Gateway returns HTTP 429 (Too Many Requests). If the client application is designed to treat 429 as a normal, retryable condition rather than a fatal error, the user experience improves because the client can back off and retry instead of surfacing raw failures. This approach prevents one noisy client from degrading service for other users and reduces the overall error rate.

Option B addresses the root cause (excess requests from a particular client) by enforcing throttling at the API Gateway layer and improving client behavior when throttled.

Option A focuses on client-side retries and error handling but does not enforce fairness or prevent a single client from consuming disproportionate API capacity. Without throttling, the same client can continue to overload downstream Lambda and DynamoDB capacity and still cause errors for everyone.

Option C (API caching) does not help for PUT requests because PUT operations are writes and are generally not cacheable in a way that reduces write load. Caching is mainly beneficial for GET responses and read-heavy workloads.

Option D (reserved concurrency) can protect Lambda capacity for critical functions, but it can also cause throttling at the Lambda layer and does not directly address controlling the request rate per API key. It also does not fix the issue at the API front door where the noisy client should be managed.

Therefore, implementing API Gateway throttling via a usage plan and ensuring clients handle 429 responses properly is the best recommendation.

The data scientists’ EC2 instances in a separate account must access S3 data in a controlled way that satisfies the policy: only authorized networks may access the IoT data. “Authorized networks” in this context means traffic must originate from approved VPCs and must not traverse the public internet.

First, traffic from the EC2 instances to S3 must stay on the AWS network without using public endpoints. A gateway VPC endpoint for S3 in the data scientists’ VPC (option A) allows EC2 instances in that VPC to reach S3 over private connectivity. When using a gateway VPC endpoint, the route tables of the subnets associated with the endpoint automatically route S3 traffic through the endpoint. This ensures that access to S3 is from an authorized network (the VPC) instead of from the public internet.

Second, access must be constrained such that only calls made through the intended S3 access mechanism are allowed. The data lake bucket already has an S3 access point. S3 access points can be restricted by policy and can be referenced in bucket policies through the s3:DataAccessPointArn condition key. By using an S3 bucket policy that allows s3:GetObject only when s3:DataAccessPointArn matches the expected access point ARN (option E), the company can enforce that all valid access uses the approved access point configuration. Combined with the access point’s own network configuration and the VPC endpoint, access is limited to authorized networks.

Option B, blocking public access on the access point, is a good general security practice but does not by itself guarantee that access is restricted to authorized VPC networks. The main enforcement must be through VPC endpoints and bucket/access point policies.

Option C denies s3:GetObject when s3:DataAccessPointArn is a valid access point ARN, which is the opposite of what is required. This would prevent intended access via the access point rather than allow it.

Option D is not correct because gateway VPC endpoints for S3 are configured via endpoint associations with route tables, not by directly routing to an access point in the route table. Traffic to S3 is routed to the VPC endpoint, and the endpoint plus S3 policies determine access.

Therefore, creating a gateway VPC endpoint for S3 in the data scientists’ VPC (option A) and using a bucket policy condition on s3:DataAccessPointArn to allow access only through the authorized access point (option E) together meet the requirement of secure, network-restricted access from authorized networks.

The company should use AWS Amplify to create a static website for uploads of media files. The company should use Amplify Hosting to serve the website through Amazon CloudFront. The company should use Amazon S3 to store the uploaded media files. The company should use Amazon Cognito to authenticate users. This solution will meet the requirements with the least operational overhead because AWS Amplify is a complete solution that lets frontend web and mobile developers easily build, ship, and host full-stack applications on AWS, with the flexibility to leverage the breadth of AWS services as use cases evolve. No cloud expertise needed1. By using AWS Amplify, the company can refactor the application to a serverless architecture that reduces operational complexity and costs. AWS Amplify offers the following features and benefits:

Amplify Studio: A visual interface that enables you to build and deploy a full-stack app quickly, including frontend UI and backend.

Amplify CLI: A local toolchain that enables you to configure and manage an app backend with just a few commands.

Amplify Libraries: Open-source client libraries that enable you to build cloud-powered mobile and web apps.

Amplify UI Components: Open-source design system with cloud-connected components for building feature-rich apps fast.

Amplify Hosting: Fully managed CI/CD and hosting for fast, secure, and reliable static and server-side rendered apps.

By using AWS Amplify to create a static website for uploads of media files, the company can leverage Amplify Studio to visually build a pixel-perfect UI and connect it to a cloud backend in clicks. By using Amplify Hosting to serve the website through Amazon CloudFront, the company can easily deploy its web app or website to the fast, secure, and reliable AWS content delivery network (CDN), with hundreds of points of presence globally. Byusing Amazon S3 to store the uploaded media files, the company can benefit from a highly scalable, durable, and cost-effective object storage service that can handle any amount of data2. By using Amazon Cognito to authenticate users, the company can add user sign-up, sign-in, and access control to its web app with a fully managed service that scales to support millions of users3.

The other options are not correct because:

Using AWS Application Migration Service to migrate the application server to Amazon EC2 instances would not refactor the application or accelerate development. AWS Application Migration Service (AWS MGN) is a service that enables you to migrate physical servers, virtual machines (VMs), or cloud servers from any source infrastructure to AWS without requiring agents or specialized tools. However, this would not address the challenges of overutilization and data uploads failures. It would also not reduce operational overhead or costs compared to a serverless architecture.

Creating a static website for uploads of media files and using AWS AppSync to create an API would not be as simple or fast as using AWS Amplify. AWS AppSync is a service that enables you to create flexible APIs for securely accessing, manipulating, and combining data from one or more data sources. However, this would require more configuration and management than using Amplify Studio and Amplify Hosting. It would also not provide authentication features like Amazon Cognito.

Setting up AWS IAM Identity Center (AWS Single Sign-On) to give users the ability to sign in to the application would not be as suitable as using Amazon Cognito. AWS Single Sign-On (AWS SSO) is a service that enables you to centrally manage SSO access and user permissions across multiple AWS accounts and business applications. However, this service is designed for enterprise customers who need to manage access for employees or partners across multiple resources. It is not intended for authenticating end users of web or mobile apps.

Create an S3 Bucket:

Navigate to Amazon S3 in the AWS Management Console and create a new S3 bucket to store the game files.Enable static website hosting on this bucket.

Upload Game Files:

Upload the 5 GB game release package to the S3 bucket. Ensure that the files are publicly accessible if required for download.

Configure Amazon Route 53:

Set up a new domain or subdomain in Amazon Route 53 and point it to the S3 bucket. This allows users to access the game files using a custom URL.

Use Amazon CloudFront:

Create a CloudFront distribution with the S3 bucket as the origin. CloudFront is a content delivery network (CDN) that caches content at edge locations worldwide, improving download performance and reducing latency for users regardless of their location.

Publish the Download URL:

Use the CloudFront distribution URL as the download link for users to access the game files. CloudFront will handle the efficient distribution and caching of the content.

This solution leverages the scalability of Amazon S3 and the performance benefits of CloudFront to provide an optimal download experience for users globally while minimizing costs.

References

Amazon CloudFront Documentation

Amazon S3 Static Website Hosting

Answer B (AWS Global Accelerator) best meets both requirements: fastest failover from application failures and the lowest response time for global users. Global Accelerator is designed for global, latency-sensitive, highly available applications. It uses static Anycast IP addresses so users enter the AWS network at the nearest edge location, then traffic is carried over the AWS global backbone to the closest healthy regional endpoint (in this case, each Region’s ALB). It continuously performs health checks and can rapidly shift traffic away from unhealthy endpoints, which helps “handle application failures quickly” without being constrained by DNS caching behavior.

Why the other options are less suitable:

C (Route 53 latency-based routing + health checks) can improve routing and failover, but it is still DNS-based. DNS-based failover can be delayed by DNS resolver caching and TTL propagation, meaning failover may not be as fast or as consistent as Global Accelerator for rapid recovery.

A (CloudFront with origin group) can provide origin failover and performance benefits mainly for cacheable or edge-optimizable content. For a payment processing API (typically dynamic and not cache-friendly), CloudFront is not the primary service for achieving the least response time globally with fast multi-Region failover compared with Global Accelerator’s Anycast + edge routing model.

D (AWS Cloud WAN) addresses private network connectivity and centralized network policy across Regions/VPCs. It does not directly optimize public user-to-application latency or provide the same kind of edge-based, Anycast-driven rapid failover for internet-facing API endpoints.

A is correct because the agent performs autonomous workload actions without user interaction. In OAuth terms, this maps to the client credentials grant, where the workload authenticates as itself rather than on behalf of an interactive user. Amazon Bedrock AgentCore Identity is designed for identity and credential management for AI agents and automated workloads, including workload identities, credential providers, and auditability. AWS documentation describes AgentCore Identity as supporting agent identities through workload identities and states that client credentials are used when an agent authenticates directly with each system using its own identity and preconfigured permissions. IAM users with long-lived keys are a weaker security design. Authorization code is user-delegated and interactive, so it does not fit a no-user-interaction autonomous agent workflow.

https://aws.amazon.com/step-functions/use-cases/

The correct solution is to use an Amazon S3 File Gateway to store the copy of the SMB file share on AWS. An S3 File Gateway enables on-premises applications to store and access objects in Amazon S3 using the SMB protocol. The S3 File Gateway can also be accessed from AWS using the SMB protocol, which provides the ability to use the data from either the data center or AWS if a disaster occurs. The S3 File Gateway supports tiering of data to different S3 storage classes based on the file type. This allows the company to optimizethe storage costs by using S3 Standard-Infrequent Access (S3 Standard-IA) for the metadata files, which are rarely accessed but must be available within 5 minutes, and S3 Glacier Deep Archive for the image files, which are the lowest-cost storage class and suitable for long-term retention of data that is rarely accessed. This solution is the most cost-effective because it does not require any additional hardware, software, or replication services.

The other solutions are incorrect because they either use more expensive or unnecessary services or components, or they do not meet the requirements. For example:

Solution A is incorrect because it uses AWS Outposts with Amazon S3 storage, which is a very expensive and complex solution for the scenario in the question. AWS Outposts is a service that extends AWS infrastructure, services, APIs, and tools to virtually any data center, co-location space, or on-premises facility. It is designed for customers who need low latency and local data processing. Amazon S3 storage on Outposts provides a subset of S3 features and APIs to store and retrieve data on Outposts. However, this solution does not provide SMB access to the data on Outposts, which requires a Windows EC2 instance on Outposts as a file server. This adds more cost and complexity to the solution, and it does not provide the ability to access the data from AWS if a disaster occurs.

Solution B is incorrect because it uses Amazon FSx File Gateway and Amazon FSx for Windows File Server Multi-AZ file system that uses SSD storage, which are both more expensive and unnecessary services for the scenario in the question. Amazon FSx File Gateway is a service that enables on-premises applications to store and access data in Amazon FSx for Windows File Server using the SMB protocol. Amazon FSx for Windows File Server is a fully managed service that provides native Windows file shares with the compatibility, features, and performance that Windows-based applications rely on. However, this solution does not meet the requirements because it does not provide the ability to use different storage classes for the metadata files and image files, and it does not provide the ability to access the data from AWS if a disaster occurs. Moreover, using a Multi-AZ file system that uses SSD storage is overprovisioned and costly for the scenario in the question, which involves rarely accessed data that must be available within 5 minutes.

Solution D is incorrect because it uses an S3 File Gateway that uses S3 Standard-IA for both the metadata files and image files, which is not the most cost-effective solution for the scenario in the question. S3 Standard-IA is a storage class that offers high durability, availability, and performance for infrequently accessed data. However, it is more expensive than S3 Glacier Deep Archive, which is the lowest-cost storage class and suitable for long-term retention of data that is rarely accessed. Therefore, using S3 Standard-IA for the image files, which are likely to be larger and more numerous than the metadata files, is not optimal for the storage costs.

What is S3 File Gateway?

Using Amazon S3 storage classes with S3 File Gateway

Accessing your file shares from AWS

Set Up EventBridge Event Bus:

Step 1: Open the Amazon EventBridge console and create a custom event bus. This bus will be used to handle user deletion events.

Step 2: Name the event bus appropriately (e.g.,user-deletion-bus).

Post Events on User Deletion:

Step 1: Modify the central user service to post an event to the custom EventBridge event bus whenever a user is deleted.

Step 2: Ensure the event includes relevant details such as the user ID and any other necessary metadata.

Create EventBridge Rules for Microservices:

Step 1: For each microservice that needs to delete user data, create a new rule in EventBridge that triggers on the user deletion event.

Step 2: Define the event pattern to match the user deletion event. This pattern should include the event details posted by the central user service.

Invoke Microservice Logic:

Step 1: Configure the EventBridge rule to invoke a target, such as an AWS Lambda function, which contains the logic to delete the user data from the microservice ' s data store.

Step 2: Each microservice should have its Lambda function or equivalent logic to handle the deletion of user data upon receiving the event.

Using Amazon EventBridge ensures a scalable, reliable, and decoupled approach to handle the deletion of user data across multiple microservices. This setup allows each microservice to independently process user deletion events without direct dependencies on other services.

References

AWS EventBridge Documentation

DynamoDB Streams and AWS Lambda Triggers

Implementing the Transactional Outbox Pattern with EventBridge Pipes​(AWS Documentation)​​(Amazon Web Services, Inc.)​​(Amazon Web Services, Inc.)​​(AWS Documentation)​​(AWS Cloud Community)​.

AWS Cost Anomaly Detection is a feature of the AWS Cost Management suite that leverages machine learning to enable continuous monitoring of your AWS costs and usage, allowing you to identify unexpected and abnormal spending1. You can create cost monitors that evaluate specific AWS services, member accounts, cost allocation tags, orcost categories based on your AWS account structure2. You can also configure alert subscriptions that notify you when a cost monitor detects an anomaly that meets your threshold2. In this case, you can create a cost monitor with a monitor type of AWS Service and apply a filter of Amazon EC2 to track the EC2 usage as a metric. You can then configure an alert subscription to notify the architecture team if the usage is 10% more than the average usage for the last 30 days, which is the anomaly detection period used by AWS Cost Anomaly Detection3.

Ais the correct best-practice approach:

Activate cost allocation tags inmanagement/payer account.

Enable AWS Cost and Usage Report (CUR) to dump usage to S3.

Query CUR withAmazon Athena, and visualize withAmazon QuickSight.

Other options either duplicate CURs (C, D) or misuse tools (B: CloudWatch is not a cost analysis tool).

The best solution is to store the database credentials for both environments in AWS Secrets Manager with distinct key entry for the QA environment and the production environment. AWS Secrets Manager is a web service that can securely store, manage, and retrieve secrets, such as database credentials. AWS Secrets Manager also supports automatic rotation of secrets by using Lambda functions or built-in rotation templates. By storing the database credentials for both environments in AWS Secrets Manager, the company can avoid exposing credentials within application code and rotate passwords automatically. By providing a reference to the Secrets Manager key as an environment variable for the Lambda functions, the company can easily access the credentials from the code by using the AWS SDK. This solution meets all the requirements of the company.

This solution allows developers to quickly launch resources using pre-approved configurations and instance types, while also ensuring that the resources launched comply with the company ' s architectural patterns. This can help reduce data transfer and compute costs associated with the resources. Using AWS Service Catalog also allows the company to control access to the approved configurations and resources through the use of IAM roles, while also allowing developers to quickly provision resources without negatively affecting their ability to perform their tasks.

it ' s the one that handles failover while B (the one shown as the answer today) it almost the same but does not handle failover.

The requirement is to estimate total cost of ownership before migration. This includes comparing current on-premises costs with projected AWS costs for compute, storage, databases, and related services, and producing a consolidated business-focused TCO view.

AWS Migration Evaluator is the AWS service specifically designed to help customers build data-driven business cases for migrating workloads to AWS. Migration Evaluator collects inventory and utilization data from on-premises environments by using a collector. It then allows architects to model different migration scenarios, map on-premises resources to AWS services such as Amazon EKS managed node groups and Amazon RDS for PostgreSQL, and generate reports that estimate costs, savings, and TCO impacts. The Quick Insights report provides a summarized, executive-level view of the estimated AWS costs, on-premises costs, and projected savings, which directly satisfies the requirement.

Option B is not sufficient because AWS DMS assessment reports focus on database migration feasibility and compatibility, not full workload TCO. The AWS Pricing Calculator can estimate AWS service costs, but it does not automatically incorporate on-premises cost data or provide a consolidated TCO comparison without significant manual effort.

Option C is incorrect because AWS Application Migration Service focuses on rehosting servers and testing migrations. It does not produce comprehensive TCO reports comparing on-premises and AWS environments for container platforms like EKS and managed databases.

Option D is incorrect because the AWS Cloud Economics Center and Cloud Value Framework provide conceptual guidance and methodology, not a workload-specific TCO report. AWS Cost and Usage Reports are used to analyze existing AWS usage, not to estimate costs for workloads that have not yet been migrated.

Therefore, using Migration Evaluator to collect on-premises data, model the target EKS and RDS architecture, and export a Quick Insights TCO report is the correct solution.

The company should create a second target group that is referenced by the ALB. The company should deploy the new logic to EC2 instances in this new target group. The company should update the ALB listener rule to use weighted target groups. The company should configure ALB target group stickiness. This solution will meet the requirements because weighted target groups are a feature that enables you to distribute traffic across multiple target groups using a single listener rule. You can specify a weight for each target group, which determines the percentage of requests that are routed to that target group. For example, if you specify two target groups, each with a weight of 10, each target group receives half the requests1. By creating a second target group and deploying the new logic to EC2 instances in this new target group, the company can have two versions of its business logic running in parallel. By updating the ALB listener rule to use weighted target groups,the company can control how much traffic is sent to each version. By configuring ALB target group stickiness, the company can ensure that a customer uses the same version of the business logic during the testing window. Target group stickiness is a feature that enables you to bind a user’s session to a specific target within a target group for the duration of the session2.

The other options are not correct because:

Creating a second ALB and deploying the new logic to a set of EC2 instances in a new Auto Scaling group would not be as cost-effective or simple as using weighted target groups. A second ALB would incur additional charges and require more configuration and management. Updating the Route 53 record to use weighted routing would not ensure that a customer uses the same version of the business logic during the testing window, as DNS caching could affect how requests are routed.

Creating a new launch configuration for the Auto Scaling group and replacing it on the Auto Scaling group would not allow for gradual traffic shifting between versions. A launch configuration is a template that an Auto Scaling group uses to launch EC2 instances. You can specify information such as the AMI ID, instance type, key pair, security groups, and block device mapping for your instances3. However, replacing the launch configuration on an Auto Scaling group would affect all instances in that group, not just 10% of customers.

Creating a second Auto Scaling group and changing the ALB routing algorithm to least outstanding requests (LOR) would not allow for controlled traffic shifting between versions. A second Auto Scaling group would require more configuration and management. The LOR routing algorithm is a feature that enables you to route traffic based on how quickly targets respond to requests. The load balancer selects a target from the target group with the fewest outstanding requests4. However, this algorithm does not take into account customer sessions or weights.

An SCP at a lower level can ' t add a permission after it is blocked by an SCP at a higher level. SCPs can only filter; they never add permissions. SO you need to create a new OU for the new account assign an SCP, and move the root SCP to Production OU. Then move the new account to production OU when AWS config is done.

https://docs.aws.amazon.com/compute-optimizer/latest/APIReference/API_ExportLambdaFunctionRecommendations.html

Amazon ECS on Fargateis ideal forevent-driven, long-running jobswith minimal management. Combine S3event notificationswithEventBridge rulesto trigger a Fargate task per upload.

Using Fargate with EventBridge

Taking a snapshot of the EBS volume using Amazon Data Lifecycle Manager (DLM) will meet the requirements because it allows you to create a backup of the volumewithout the need to access the instance or its SSH key pair. Additionally, DLM allows you to schedule the backups to occur at specific intervals and also enables you to copy the snapshots to an S3 bucket. This approach will not impact the running application as the backup is performed on the EBS volume level.

This is a data residency and preventive governance problem. EU customer uploads must be routed to EU endpoints, processed by EU workloads, and stored only in EU Regions. Separate EU and non-EU stacks avoid commingling workloads and make residency controls enforceable. Route 53 geolocation routing is appropriate because the routing decision is based on user location, not latency. The critical control is an Organizations SCP that denies resource creation outside approved Regions. AWS documents region-deny patterns using the aws:RequestedRegion condition and AWS Control Tower Region deny controls to prevent actions outside governed Regions. AWS Config is detective, not preventive, so option C is too late. Latency-based routing can send EU users outside the EU, and a single global API design weakens the sovereignty boundary.

A is the correct answer because it uses Amazon DocumentDB (with MongoDB compatibility) as a key-value database that can scale based on demand and supports encryption in transit and at rest. Amazon DocumentDB is a fully managed document database service that is designed to be compatible with the MongoDB API. It is a NoSQL database that is optimized for storing, indexing, and querying JSON data. Amazon DocumentDB supports encryption in transit using TLS and encryption at rest using AWS Key Management Service (AWS KMS). Amazon DocumentDB also supports provisioned IOPS volumes that can scale up to 64 TiB of storage and 256,000 IOPS per cluster. To connect to Amazon DocumentDB, you can use the instance endpoint, which connects to a specific instance in the cluster, or the cluster endpoint, which connects to the primary instance or one of the replicas in the cluster. Using the cluster endpoint is recommended for high availability and load balancing purposes. References:

https://docs.aws.amazon.com/documentdb/latest/developerguide/what-is.html

https://docs.aws.amazon.com/documentdb/latest/developerguide/security.encryption.html

https://docs.aws.amazon.com/documentdb/latest/developerguide/limits.html

https://docs.aws.amazon.com/documentdb/latest/developerguide/connecting.html

This solution will meet the requirements with the least operational overhead because it allows the company to modify the existing environment ' s capacity configuration, so it becomes a load-balanced environment type. By selecting all availability zones, the company can ensure that the application is running in multiple availability zones, which can help to improve the availability and scalability of the application. The company can also add a scale-out rule that will run if the average CPU utilization is over 85% for 5 minutes, which can help to mitigate the performance issues. This solution does not require creating new Elastic Beanstalk environments or rebuilding the existing one, which reduces the operational overhead.

You can refer to the AWS Elastic Beanstalk documentation for more information on how to use this service:https://aws.amazon.com/elasticbeanstalk/You can refer to the AWS documentation for more information on how to use autoscaling:https://aws.amazon.com/autoscaling/

The CAPABILITY_NAMED_IAM capability is required when creating or updating CloudFormation stacks that contain IAM resources with custom names. This capability acknowledges that the template mightcreate IAM resources that have broad permissions or affect other resources in the AWS account. The stack set’s template contains an IAM role that has a custom name, so this capability is needed. Enabling the new Regions in all relevant accounts is also necessary to deploy the stack set across multiple Regions and accounts.

Option B is incorrect because the Service Quotas console is used to view and manage the quotas for AWS services, not for CloudFormation stacks. The number of stacks per Region per account is not a service quota that can be increased.

Option C is incorrect because the SELF_MANAGED permissions model is used when the administrator wants to retain full permissions to manage stack sets and stack instances. This model does not affect the creation of the stack set or the requirement for the CAPABILITY_NAMED_IAM capability.

Option D is incorrect because an administration role ARN is optional when creating a stack set. It is used to specify a role that CloudFormation assumes to create stack instances in the target accounts. It does not affect the creation of the stack set or the requirement for the CAPABILITY_NAMED_IAM capability.

1: AWS CloudFormation stack sets

2: Acknowledging IAM resources in AWS CloudFormation templates

3: AWS CloudFormation stack set permissions

https://aws.amazon.com/blogs/mt/self-service-vpcs-in-aws-control-tower-using-aws-service-catalog/

https://docs.aws.amazon.com/vpc/latest/tgw/tgw-transit-gateways.htmlhttps://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-resource-ec2-transitgatewayattachment.html

The best option is to use an Auto Scaling group, a Network Load Balancer, Amazon Route 53, and Amazon DynamoDB to create a scalable, highly available, and low-latency online game application. An Auto Scaling group can automatically adjust the number of EC2 instances based on the demand and traffic in each Region. A Network Load Balancer can distribute the incoming traffic across the EC2 instances and handle millions of requests per second. Amazon Route 53 can use latency-based routing to direct the users to the Region that provides the best performance. Amazon DynamoDB global tables can replicate the game metadata across multiple Regions, ensuring consistency and availability of thedata. This approach minimizes the operational overhead and cost, as it leverages fully managed services and avoids custom logic or replication.

Option A is not optimal because using an EC2 Spot Fleet can introduce the risk of losing the EC2 instances if the Spot price exceeds the bid price, which can affect the availability and performance of the game. Using AWS Global Accelerator can improve the network performance, but it is not necessary if Amazon Route 53 can already route the users to the closest Region. Using Amazon RDS for MySQL can store the game metadata, but it requires setting up read replicas and managing the replication lag across Regions, which can increase the complexity and cost.

Option B is not optimal because using geoproximity routing can direct the users to the Region that is geographically closer, but it does not account for the network latency or performance. Using MySQL databases on EC2 instances can store the game metadata, but it requires managing the EC2 instances, the database software, the backups, the patches, and the replication across Regions, which can increase the operational overhead and cost.

Option D is not optimal because using EC2 Global View is not a valid service. Using a custom DNS server on an EC2 instance can redirect the user to the Region that provides the lowest latency, but it requires developing and maintaining the custom logic, as well as managing the EC2 instance, which can increase the operational overhead and cost. Using Amazon Aurora global database can store the game metadata, but it is more expensive and complex than using Amazon DynamoDB global tables.

Auto Scaling groups

Network Load Balancer

Amazon Route 53

Amazon DynamoDB global tables

This solution will improve the reliability of the application by addressing the issues of scalability, availability, and performance. Containerizing the application will make it easier to deploy and manage on AWS. Migrating the application to an Amazon ECS cluster will allow the application to run on a fully managed container orchestration service. Using the AWS Fargate launch type for the tasks that host the application will enable the application to run on serverless compute engines that are automatically provisioned and scaled by AWS. Creating an Amazon EFS file system for the static content will provide a scalable and shared storage solution that can be accessed by multiple containers. Mounting the EFS file system to each container will eliminate the need to copy the static content to each EBS volume and ensure that the content is always up to date. Configuring AWS Application Auto Scaling on the ECS cluster will enable the application to scale up and down based on demand or a predefined schedule. Setting the ECS service as a target for the ALB will distribute the incoming requests across multiple tasks in the ECS cluster and improve the availability and fault tolerance of the application. Migrating the database to Amazon Aurora MySQL Serverless v2 with a reader DB instance will provide a fully managed, compatible, and scalable relational database service that can handle high throughput and concurrent connections. Using a reader DB instance will offload some of the read load from the primary DB instance and improve the performance of the database.

The correct answer is B.

B. This solution meets the requirements because it uses AWS Cost Anomaly Detection, which is a feature of AWS Cost Management that uses machine learning to identify and alert on anomalous spend and usage patterns. By configuring a monitor type of AWS Service and applying a filter of Amazon EC2, the solution can track the EC2 usage as a metric across the organization’s accounts.By configuring an alert subscription with a threshold of 10%, the solution can notify the architecture team via email or AmazonSNS if the EC2 usage is more than 10% higher than the average usage for the last 30 days12

A. This solution is incorrect because it uses AWS Budgets, which is a feature of AWS Cost Management that helps to plan and track costs and usage. However, AWS Budgets does not support usage type of EC2 running hours as a budget type. The only supported usage types are Amazon S3 storage, Amazon EC2 RI utilization, and Amazon EC2 RI coverage. Moreover, AWS Budgets does not support setting the budget amount based on the reported average usage from AWS Cost Explorer.The budget amount has to be a fixed or variable value34

C. This solution is incorrect because it uses AWS Trusted Advisor, which is a feature of AWS Premium Support that provides recommendations to follow best practices for cost optimization, security, performance, and fault tolerance. However, AWS Trusted Advisor does not support configuring custom alerts based on EC2 usage or average usage for the last 30 days.The only supported alerts are based on predefined checks and thresholds that are applied to all services and resources in the account56

D. This solution is incorrect because it uses Amazon Detective, which is a service that helps to analyze and visualize security data to investigate potential security issues. However, Amazon Detective does not support configuring EC2 usage anomaly alerts basedon average usage for the last 30 days.The only supported alerts are based on GuardDuty findings and other security-related events that are detected by machine learning models78

Option D is correct because the company needs gradual release control and reduced infrastructure management. SageMaker Serverless Inference removes the need to select, run, and scale inference instances, which directly reduces operational overhead compared with EC2-hosted Python inference or manually managed endpoints. API Gateway canary deployments allow a defined percentage of API traffic to be routed to a newer deployment while the rest remains on the stable production deployment. AWS documentation states that API Gateway canary settings include a percentage of API traffic and stage variable overrides, which can direct canary traffic to the new backend. Options B and C require custom Lambda routing logic, which increases operational responsibility. Option A incorrectly treats a multi-model endpoint as the cleanest version-rollout mechanism for this canary API requirement.

https://docs.aws.amazon.com/elasticloadbalancing/latest/application/load-balancer-access-logs.html

Adding an accelerator in AWS Global Accelerator will enable improving the performance of the APIs for local and global users1. AWS Global Accelerator is a service that uses the AWS global network to route traffic to the optimal regional endpoint based on health, client location, and policies1. Configuring the ALB as the origin will enable connecting the accelerator to the ALB that exposes the APIs2. AWS Global Accelerator supports non-standard REST methods such as LINK, UNLINK, LOCK, and UNLOCK3.

The correct solution is to use the authenticate-oidc action for the ALB listener rule and configure it with the details of the AWS Directory Service for Microsoft Active Directory directory. This way, the ALB can use OpenID Connect (OIDC) to authenticate users against the directory and grant them access to the intranet web application. The app client in the directory is used to register the ALB as an OIDC client and provide the necessary credentials and endpoints. The callback URL is the URL that the ALB redirects the user to after a successful authentication. This solution does not require any additional services or roles, and it leverages the existing directory accounts for all users.

The other solutions are incorrect because they either use the wrong action for the ALB listener rule, or they involve unnecessary or incompatible services or roles. For example:

Solution B is incorrect because it uses Amazon Cognito user pool, which is a separate user directory service that does not integrate with AWS Directory Service for Microsoft Active Directory. To use this solution, the company would have to migrate or synchronize their users from the directory to the user pool, which is not required by the question. Moreover, the authenticate-cognito action for the ALB listener rule only works with Amazon Cognito user pools, not with federated identity providers (IdPs) that have metadata from the directory.

Solution C is incorrect because it uses IAM as an identity provider (IdP), which is not compatible with AWS Directory Service for Microsoft Active Directory. IAM can only be used as an IdP for web identity federation, which allows users to sign in with social media or other third-party IdPs, not with Active Directory. Moreover, the authenticate-oidc action for the ALB listener rule requires an OIDC IdP, not a SAML 2.0 federation IdP, which is what IAM provides.

Solution D is incorrect because it uses AWS IAM Identity Center (AWS Single Sign-On), which is a service that simplifies the management of SSO access to multiple AWS accounts and business applications. This service is not needed for the scenario in the question, which only involves a single intranet web application. Moreover, the authenticate-cognito action for the ALB listener rule does not work with external IdPs that use SAML, such as AWS IAM Identity Center.

Authenticate users using an Application Load Balancer

What is AWS Directory Service for Microsoft Active Directory?

Using OpenID Connect for user authentication

The modernized application needs to store and serve video files up to 4 GB. Large binary content should not be stored directly in a database because it negatively affects database performance, scaling, and backup/restore times. The standard AWS pattern is to store large objects in Amazon S3 and keep only references (such as keys or URLs) in the database.

User profile data is simple, text-based, and currently lives in a single table. This is a good fit for a key-value or document-style data model such as Amazon DynamoDB, which provides fully managed horizontal scaling and predictable performance at scale.

Option B migrates the profile table to DynamoDB using AWS DMS with AWS Schema Conversion Tool (SCT). Video files are stored as objects in Amazon S3, which is designed for large, durable, and highly scalable object storage. The DynamoDB item stores only the S3 key that corresponds to each user’s video, keeping the database small and responsive. S3 can scale rapidly and independently of the database, and offloads the heavy I/O load of large objects away from the transactional store, so overall application performance is not negatively impacted by video storage or retrieval.

Option A stores the videos as base64-encoded strings in a TEXT column in a relational database. Storing multi-gigabyte objects directly in a relational database table leads to large row sizes, larger indexes, longer backup times, and degraded performance as the table grows. This does not meet the requirement to avoid affecting application performance.

Option C uses Amazon Keyspaces (for Apache Cassandra) as the profile store and S3 for video. While storing videos in S3 is correct, Keyspaces is typically chosen when there is already a Cassandra-based workload or a need for Cassandra-compatible interfaces. For a simple single-table user profile store being migrated from MySQL, DynamoDB is the more natural, managed choice and aligns better with AWS migration and modernization guidance.

Option D stores the videos as base64-encoded strings in DynamoDB items. DynamoDB has a maximum item size of 400 KB, which is far smaller than the 4 GB video size requirement. This makes option D technically infeasible.

Therefore, migrating user profile data to DynamoDB and storing video files in Amazon S3, with S3 keys stored in the corresponding DynamoDB items (option B), provides scalable video storage and protects application performance.

The issue is API latency for users outside the Region where the API Gateway Regional API endpoint is deployed. The static site is already delivered through CloudFront, but backend API calls bypass CloudFront and go directly to the Regional API endpoint. The best design is to configure API Gateway as another origin in the existing CloudFront distribution and have the application call the same application DNS name. AWS guidance for single-page applications shows CloudFront serving static S3 content and routing dynamic REST/API requests through additional origins, improving global access through CloudFront edge locations. API Gateway edge-optimized behavior also uses CloudFront to route API requests through the nearest edge location. Options A and B do not accelerate the API path. Global Accelerator is not the correct direct integration for API Gateway here.

Reference architecture -https://docs.aws.amazon.com/vpc/latest/privatelink/privatelink-access-saas.html

Note from documentation that Interface Endpoint is at client side

This solution uses a combination of AWS Resource Access Manager (RAM) and automated backups to migrate the Lambda functions and the Aurora database to theTarget account while minimizing downtime. In this solution, the Lambda function deployment package is downloaded from the Source account and used to create new Lambda functions in the Target account. The Aurora DB cluster is shared with the Target account using AWS RAM and the Target account is granted permission to clone the Aurora DB cluster, allowing for a new copy of the Aurora database to be created in the Target account. This approach allows for the data to be migrated to the Target account while minimizing downtime, as the Target account can use the cloned Aurora database while the original Aurora database continues to be used in the Source account.

Installing AWS Outposts for the applications that have data regulatory requirements or requirements for latency of single-digit milliseconds will provide a fully managed service that extends AWS infrastructure, services, APIs, and tools to customer premises1. AWS Outposts allows customers to run some AWS services locally and connect to a broad range of services available in the local AWS Region1. Using AWS Snowball Edge Compute Optimized devices to host the workloads in the factory sites will provide local compute and storage resources for locations with limited network infrastructure2. AWS Snowball Edge devices can run Amazon EC2 instances and AWS Lambda functions locally and sync data with AWS when network connectivity is available2.

Storing the database credentials in AWS Secrets Manager as a secret that is associated with an AWS Key Management Service (AWS KMS) customer managed key will enable encrypting and managing the credentials securely1. AWS Secrets Manager helps you to securely encrypt, store, and retrieve credentials for your databases and other services2. Attaching a role to each Lambda function to provide access to the secret will enable retrieving the credentials programmatically1. Restricting access to the secret and the customer managed key so that only members of the IT security team’s IAM user group can access them will enable meeting the security requirements1.

Option A, Deploy the security tool on EC2 instances in a new Auto Scaling group in the existing VPC, allows the company to use its existing security tool while still running it within the AWS environment. This ensures that all packets coming in and out of the VPC are inspected by the security tool in real time. Option D, Provision a Gateway Load Balancer for each Availability Zone to redirect the traffic to the security tool, allows for high availability within an AWS Region. By provisioning a Gateway Load Balancer for each Availability Zone, the traffic is redirected to the security tool in the event of any failures or outages. This ensures that the security tool is always available to inspect the traffic, even in the event of a failure.

The workload uses EC2 Spot Instances, which are subject to capacity availability constraints. Using a single instance type significantly increases the risk of launch failures when Spot capacity for that instance type is constrained. This directly impacts application reliability and increases user wait times.

Attribute-based instance type selection allows Auto Scaling to choose from multiple compatible instance types based on defined requirements such as vCPU count, memory, and architecture. This improves the probability that Spot capacity is available at launch time because Auto Scaling can select from a broader pool of instance types instead of a single one.

Option D correctly applies attribute-based instance type selection within a new version of the existing launch template. Auto Scaling groups are designed to work with launch templates, and updating the template version is the modern and recommended approach. This change increases Spot placement flexibility and reliability without altering application behavior or introducing additional infrastructure.

Option A is incorrect because launch configurations are an older construct and do not support attribute-based instance type selection. They are also not recommended for new or updated Auto Scaling designs.

Option B does not address the root cause of Spot launch failures. Simply using a larger instance type may actually reduce available Spot capacity and worsen reliability.

Option C increases the number of placement groups, but placement groups do not solve Spot capacity shortages. They influence instance placement for networking or latency characteristics, not Spot availability across instance types.

Therefore, updating the launch template to use attribute-based instance type selection is the most effective way to improve reliability for a Spot-based Auto Scaling workload.

Deploying the application on Amazon EKS with managed node groups simplifies the operational overhead of managing the Kubernetes cluster. Running the web servers and application as Kubernetes deployments ensures that the desired number of pods are always running and can scale up or down as needed. Storing the frontend web server session data in an Amazon DynamoDB table provides a fast, scalable, and durable storage option that can be accessed across multiple Availability Zones. Creating an Amazon EFS volume that all applications will mount at the time of deployment allows the application to share data that is frequently accessed between the web and application tiers. References:

https://docs.aws.amazon.com/eks/latest/userguide/managed-node-groups.html

https://docs.aws.amazon.com/eks/latest/userguide/deployments.html

https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/Introduction.html

https://docs.aws.amazon.com/efs/latest/ug/mounting-fs.html

The best solution is to turn on the Concurrency Scaling feature for the Amazon Redshift cluster. This feature allows the cluster to automatically add additional capacity to handle bursts of read queries without affecting the performance of write queries. The additional capacity is transparent to the users and is billed separately based on the usage. This solution meets the business requirements of servicing read and write queries at all times andis also cost-effective compared to the other options, which involve provisioning additional resources or resizing the cluster. References: Amazon Redshift Documentation, Concurrency Scaling in Amazon Redshift

Option C is correct because the business process is a multi-step workflow that needs state tracking: order received, package label printed, package scanned, and order marked as shipped. AWS Step Functions is the correct serverless orchestration service for coordinating these steps, while AWS Lambda performs event-driven work and DynamoDB stores order state. This removes the email-based manual workflow and avoids continuously running EC2 workers. AWS documentation describes Step Functions as a workflow service that can include human-in-the-loop and callback-style patterns, making it suitable for business processes that wait for external completion signals. AWS Batch is for batch compute jobs, not warehouse process orchestration. SQS with polling every 5 minutes is less direct and still introduces EC2 in option B. EFS plus instances is not a serverless modernization pattern.

Provision an Aurora Replica in a different Region will meet the requirement of the application being able to recover to a separate AWS Region in the event of an application failure, and no data can be lost, with the least amount of operational overhead.

Option B is correct because a first migration workload should be low risk, relatively independent, and useful for building team experience before moving business-critical systems. An internal marketing portal is a better candidate than claims processing, trading, or compliance workloads because failure has lower operational impact. Rehosting with AWS Application Migration Service supports a fast lift-and-shift migration, which is appropriate when an on-premises contract expires in less than 6 months. Adding AWS Systems Manager Patch Manager also establishes operational discipline after migration. Option A introduces unnecessary re-architecture too early. Option C selects a latency-sensitive trading workload, which is too risky as a first migration. Option D adds database conversion complexity and compliance risk, which slows the migration and increases the chance of failure.

C: Multipart uploads can leave incomplete parts behind, which incur storage costs. Expiring them after 7 days minimizes waste and saves cost.

E: Since objects are infrequently accessed after 180 days, transitioning toS3 Standard-IAis cost-effective, especially for large files > 128 KB (your 100 MB+ files qualify).

Ais for shifting download cost, not reducing your S3 storage expenses.

Bhelps with upload speed butincreases cost.

Dis too aggressive; Glacier is not suited for access patterns within the first few days.