Compute
Ways to run code on AWS, from raw virtual machines to fully abstracted platforms. Higher up the list = more control; lower = less to manage.
| Service | What it is | Choose when | Key distinction |
|---|---|---|---|
| Amazon EC2 | Virtual machines you fully control | You need OS-level control or a specific configuration | Most flexible, most to manage; the baseline everything else abstracts away |
| Amazon Lightsail | Bundled VPS at a fixed monthly price | Small, predictable workload; simplest possible server | Fixed bundle, hides EC2’s knobs |
| AWS Elastic Beanstalk | PaaS — upload code, it provisions the infra | Deploy a web app fast but keep the underlying EC2 | AWS manages resources you still own and can tune |
| AWS Batch | Managed batch job runner | Long-running compute jobs in a queue | Provisions capacity per job, then tears it down |
| AWS Local Zones | Region capability in a metro area | Single-digit-ms latency to a specific city | AWS-owned facility on the public internet |
| AWS Wavelength | Compute inside 5G carrier networks | Ultra-low latency to mobile users | Lives in the telecom network |
| AWS Outposts | AWS hardware in your data center | Data must stay on premises | You host AWS-owned racks |
- Serverless compute (AWS Lambda, AWS Fargate) lives under Serverless.