Basic Information
| Field | Details |
|---|---|
| English name | AWS Batch |
| Full name | AWS Batch |
| Chinese description | 托管批处理计算服务 |
| Japanese description | AWS Batch（バッチ処理） |
| Exam frequency | ⭐⭐⭐ |
| Often confused with | Lambda / Step Functions / ECS |
In one sentence
Managed scheduling and execution for batch jobs that can queue, retry, and run on suitable compute capacity.
Key points
- Batch manages job queues, scheduling, dependencies, retries, and compute environments.
- It fits non-interactive, retryable work such as data processing, rendering, simulation, and scientific jobs.
- Jobs can run on EC2 or Fargate capacity and may use Spot when interruption is acceptable.
Exam takeaway
Lambda is for short event-driven functions; Batch is better for longer or resource-intensive queued jobs.
Update: HPC job boundaries
- A job definition describes work, a job queue orders it, and a compute environment supplies capacity.
- Array jobs fit many independent tasks; multi-node parallel jobs coordinate work across nodes.
- EC2-based MNP does not support Spot compute environments, and Fargate does not support MNP; Spot support does not apply to every Batch mode.