Amazon EKS is intended to run Kubernetes-based container workloads using an AWS-managed control plane because eKS integrates managed Kubernetes cluster control plane services with AWS infrastructure; in this design the new outcome is to handle workloads requiring GPUs or specialized hardware acceleration, which calls for Accelerated computing EC2 instance rather than Amazon EKS or the earlier General purpose EC2 instance.

Spot interruption handling is intended to design flexible EC2 workloads to withstand capacity reclamation because Spot capacity may be interrupted by AWS, so tasks should checkpoint or retry; in this design the new outcome is to handle workloads requiring GPUs or specialized hardware acceleration, which calls for Accelerated computing EC2 instance rather than Spot interruption handling or the earlier General purpose EC2 instance.

General purpose EC2 instance primarily supports the earlier need to balance CPU memory and networking for broadly mixed workloads because general-purpose instances provide a balanced resource profile without specializing in one bottleneck, but the new task is to handle workloads requiring GPUs or specialized hardware acceleration; that is why Accelerated computing EC2 instance is preferable here.

S3 Glacier Deep Archive is intended to retain long-lived archives at very low storage cost when restoration can be slow because deep Archive is designed for extremely infrequent access and long retention; in this design the new outcome is to assess storage cost beyond the gigabyte-month charge, which calls for S3 request and retrieval pricing rather than S3 Glacier Deep Archive or the earlier AWS Storage Gateway.

AWS Storage Gateway primarily supports the earlier need to connect on-premises applications to cloud-backed storage interfaces because storage Gateway supports hybrid file volume and tape integration scenarios, but the new task is to assess storage cost beyond the gigabyte-month charge; that is why S3 request and retrieval pricing is preferable here.

EBS volume encryption is intended to protect block storage with supported at-rest encryption because encrypted EBS volumes use integrated key management for data at rest; in this design the new outcome is to assess storage cost beyond the gigabyte-month charge, which calls for S3 request and retrieval pricing rather than EBS volume encryption or the earlier AWS Storage Gateway.

EC2 Auto Scaling is intended to adjust the number of matching EC2 instances as demand changes because EC2 Auto Scaling can launch and terminate instances according to configured scaling policies; in this design the new outcome is to relaunch instances from a defined system image, which calls for Amazon EC2 AMI rather than EC2 Auto Scaling or the earlier EC2 placement groups.

EC2 placement groups primarily supports the earlier need to influence how supported instances are distributed across underlying hardware because placement groups support strategies for network performance or fault isolation, but the new task is to relaunch instances from a defined system image; that is why Amazon EC2 AMI is preferable here.

EC2 dedicated host is intended to place eligible instances on physical servers dedicated to one customer because dedicated Hosts provide visibility or control over host placement for some licensing requirements; in this design the new outcome is to relaunch instances from a defined system image, which calls for Amazon EC2 AMI rather than EC2 dedicated host or the earlier EC2 placement groups.

S3 lifecycle rule addresses a different objective, to transition or expire objects automatically based on age and criteria, because lifecycle configuration can move objects into eligible classes or delete them on schedule; the checklist instead requires the distinct outcomes to recognize block devices compared with managed network file systems; archive data with retrieval time measured in minutes to hours.

Amazon FSx for Windows File Server addresses a different objective, to provide managed SMB file shares for Windows workloads, because fSx for Windows File Server supports Windows file protocols and Active Directory integration; the checklist instead requires the distinct outcomes to recognize block devices compared with managed network file systems; archive data with retrieval time measured in minutes to hours.

Amazon EFS addresses a different objective, to mount elastic shared file storage from multiple Linux workloads, because eFS supplies managed NFS file systems that can be accessed concurrently by supported clients; the checklist instead requires the distinct outcomes to recognize block devices compared with managed network file systems; archive data with retrieval time measured in minutes to hours.

Amazon S3 Standard is intended to serve frequently accessed objects with low-latency retrieval because S3 Standard is appropriate for data requiring regular access across multiple AZs; in this design the new outcome is to archive data with retrieval time measured in minutes to hours, which calls for S3 Glacier Flexible Retrieval rather than Amazon S3 Standard or the earlier S3 Transfer Acceleration.

S3 Transfer Acceleration primarily supports the earlier need to speed long-distance S3 uploads using AWS edge networking because transfer Acceleration routes supported transfers through an accelerated endpoint, but the new task is to archive data with retrieval time measured in minutes to hours; that is why S3 Glacier Flexible Retrieval is preferable here.

S3 server-side encryption is intended to encrypt S3 object data at rest using supported key arrangements because S3 encrypts stored object data at rest and supports key-management choices; in this design the new outcome is to archive data with retrieval time measured in minutes to hours, which calls for S3 Glacier Flexible Retrieval rather than S3 server-side encryption or the earlier S3 Transfer Acceleration.

It offers more host-level control than event-driven Lambda functions; that caveat is associated more closely with Amazon EC2 for the need to operate virtual servers with explicit OS and instance management than with Amazon CloudWatch EC2 metrics for the requirement to monitor resource behavior to inform scaling and alerts.

Changing one instance type does not remove that machine as a failure point; that caveat is associated more closely with Vertical EC2 scaling for the need to increase resource capacity by changing the instance type than with Amazon CloudWatch EC2 metrics for the requirement to monitor resource behavior to inform scaling and alerts.

It is designed for functions rather than long-running self-managed operating systems; that caveat is associated more closely with AWS Lambda for the need to execute event-triggered code without managing virtual servers than with Amazon CloudWatch EC2 metrics for the requirement to monitor resource behavior to inform scaling and alerts.

It is designed for functions rather than long-running self-managed operating systems; that caveat is associated more closely with AWS Lambda for the need to execute event-triggered code without managing virtual servers than with EC2 instance store for the requirement to use temporary host-attached storage with very low latency.

It offers more host-level control than event-driven Lambda functions; that caveat is associated more closely with Amazon EC2 for the need to operate virtual servers with explicit OS and instance management than with EC2 instance store for the requirement to use temporary host-attached storage with very low latency.

Local performance does not imply durable storage across host failure; that caveat is associated more closely with Storage optimized EC2 instance for the need to run applications needing high local storage IOPS or throughput than with EC2 instance store for the requirement to use temporary host-attached storage with very low latency.

S3 bucket policy is intended to control allowed and denied access to an S3 bucket and its objects because bucket policies are resource-based JSON permissions for supported S3 actions; in this design the new outcome is to replicate eligible objects between buckets within the same AWS Region, which calls for S3 Same-Region Replication rather than S3 bucket policy or the earlier Amazon S3 One Zone-IA.

Amazon S3 Inventory is intended to generate lists of objects and associated metadata for analysis because S3 Inventory produces scheduled reports about bucket objects and properties; in this design the new outcome is to replicate eligible objects between buckets within the same AWS Region, which calls for S3 Same-Region Replication rather than Amazon S3 Inventory or the earlier Amazon S3 One Zone-IA.

Amazon S3 One Zone-IA primarily supports the earlier need to store recreatable infrequently accessed objects within one Availability Zone because one Zone-IA has a single-AZ resilience profile and lower storage cost, but the new task is to replicate eligible objects between buckets within the same AWS Region; that is why S3 Same-Region Replication is preferable here.

S3 Glacier Flexible Retrieval addresses the separate need to archive data with retrieval time measured in minutes to hours because flexible Retrieval supports lower-cost archival storage using asynchronous restore options; it is not as direct as S3 Cross-Region Replication for the actual requirement to copy eligible objects asynchronously to a bucket in a different AWS Region.

S3 server-side encryption addresses the separate need to encrypt S3 object data at rest using supported key arrangements because S3 encrypts stored object data at rest and supports key-management choices; it is not as direct as S3 Cross-Region Replication for the actual requirement to copy eligible objects asynchronously to a bucket in a different AWS Region.

Amazon FSx for Windows File Server addresses the separate need to provide managed SMB file shares for Windows workloads because fSx for Windows File Server supports Windows file protocols and Active Directory integration; it is not as direct as S3 Cross-Region Replication for the actual requirement to copy eligible objects asynchronously to a bucket in a different AWS Region.

Memory optimized EC2 instance addresses the separate need to support applications with large in-memory working sets because memory-optimized families provide higher RAM-to-vCPU ratios for memory-bound applications; it is not as direct as Amazon EC2 for the actual requirement to operate virtual servers with explicit OS and instance management.

Horizontal EC2 scaling addresses the separate need to add parallel instance capacity for a distributed application because horizontal scaling adds multiple instances that can share workload when architecture supports it; it is not as direct as Amazon EC2 for the actual requirement to operate virtual servers with explicit OS and instance management.

AWS Lambda addresses the separate need to execute event-triggered code without managing virtual servers because lambda runs functions in managed execution environments with service-defined execution constraints; it is not as direct as Amazon EC2 for the actual requirement to operate virtual servers with explicit OS and instance management.

Amazon ECR addresses the separate need to store and distribute container images for deployment because eCR is a managed container image registry integrated with AWS deployment tooling; it is not as direct as Spot interruption handling for the actual requirement to design flexible EC2 workloads to withstand capacity reclamation.

Vertical EC2 scaling addresses the separate need to increase resource capacity by changing the instance type because vertical scaling provides more CPU or RAM within a single virtual machine; it is not as direct as Spot interruption handling for the actual requirement to design flexible EC2 workloads to withstand capacity reclamation.

Memory optimized EC2 instance addresses the separate need to support applications with large in-memory working sets because memory-optimized families provide higher RAM-to-vCPU ratios for memory-bound applications; it is not as direct as Spot interruption handling for the actual requirement to design flexible EC2 workloads to withstand capacity reclamation.

EC2 user data addresses the separate need to perform initial bootstrap actions during instance launch because user data can provide launch-time configuration and installation scripts; it is not as direct as Spot interruption handling for the actual requirement to design flexible EC2 workloads to withstand capacity reclamation.