2.4Compute, Container, Database, Storage, and Network Services

New designs test picking the managed service that best fits requirements. This section organizes building blocks by definition, role, and when to choose.

2.4.1Simple and PaaS compute

• AWS App Runner: fully managed service that builds, deploys, and scales web apps/APIs automatically from a container or source. Choose it to ship a containerized web service fast with no infrastructure to manage.
• AWS Elastic Beanstalk: a PaaS that auto-provisions EC2/ELB/Auto Scaling when you upload code. Choose it to run a standard web app quickly while retaining control of the underlying stack.
• Amazon Lightsail: a simple, flat-rate VPS bundling instances/DB/LB. Choose it to stand up small sites or experiments quickly at a predictable monthly price.

2.4.2Hybrid, edge, and containers

• AWS Outposts: AWS hardware installed on-premises operated with the same APIs. Choose it to keep workloads on-site for latency or data-residency needs while retaining AWS operations.
• AWS Wavelength: deploys AWS at 5G carrier edges for ultra-low latency. Choose it when mobile devices need millisecond-level responses.
• Amazon ECS Anywhere: runs and manages ECS tasks on your own servers (on-prem, etc.). Choose it to keep the control plane in AWS while running workloads on-site.
• Amazon EKS Anywhere: builds and operates Kubernetes clusters on your own infrastructure consistently with AWS. Choose it to run EKS-compatible Kubernetes in your datacenter.
• Amazon EKS Distro: the open-source Kubernetes distribution that EKS is built on. For advanced teams assembling the same K8s themselves (most choose managed EKS/EKS Anywhere).

2.4.3Purpose-built databases

• Amazon Aurora Serverless: Aurora that auto-scales capacity with load. Choose it for intermittent or unpredictable usage where you want to avoid provisioning.
• Amazon DocumentDB: managed MongoDB-compatible document database. Choose it for JSON-document workloads using the MongoDB API.
• Amazon Keyspaces: serverless Apache Cassandra-compatible wide-column database. Choose it for large write workloads using CQL without operational overhead.
• Amazon Neptune: a graph database (property graph/RDF). Choose it when queries traverse relationships, e.g., social graphs or fraud detection.
• Amazon Timestream: a time-series database. Choose it to store and analyze large volumes of IoT sensor or operational metrics over time.

2.4.4DR, archive, hybrid storage, and delivery

• AWS Elastic Disaster Recovery (DRS): continuously replicates servers at the block level from on-prem/cloud and fails over to AWS quickly on disaster. Choose it for low-RPO/RTO server DR.
• Amazon S3 Glacier: very low-cost long-term archive storage classes. Choose it to retain rarely accessed data cheaply for compliance retention or backup.
• AWS Storage Gateway: hybrid storage bridging on-prem to S3/EBS (File/Volume/Tape gateways). Choose it to let existing apps use AWS storage transparently for migration or backup.
• AWS Global Accelerator: static Anycast IPs over the AWS backbone routing to the nearest healthy Region with low latency and instant failover. Choose it for global TCP/UDP apps that need static IPs (use CloudFront for HTTP caching/delivery).

2.4.5Section summary

• Simple/PaaS = App Runner, Elastic Beanstalk, Lightsail; hybrid = Outposts, Wavelength, ECS/EKS Anywhere
• DB by data shape = Aurora Serverless/DocumentDB/Keyspaces/Neptune/Timestream; storage = DRS, S3 Glacier, Storage Gateway