AWS Batch
Overview
- AWS Managed
Batch processing
Product - Batch processing - jobs that can run without end user interaction or can be scheduled to run as resources permit
- Lets customer define jobs and it handles compute and orchestration
- Let overcome
15 minutes
limitation of AWS Lambda - It can be dropped into many event-driven workflows for jobs which go beyond the capabilities of Lambda.
Components
- Job
- Script, Exeuctable or Docker container submitted to batch
- can reference other jobs by name or ID
- cna be dependent on the successful completion of other jobs
-
Job Definitions
- MetaData for a job
- IAM Resources
- Resource config - cpu and memory
- Environment vars and mount points
-
Job Queues
- Jobs are submitted to a queue
- Batch job resides until it is scheduled onto a compute environment.
- can associatew one or more compute environments with a job queue
- Compute Environment
- Managed Compute Environment
- Batch manages the capacity based on workolads
- You can choose EC2 isntances - On-demand or Spot Instances - size./type
- You can determine your own max spot price
- VPC - Public/Private -- requires VPC gateways
- Unmanaged Compute Environment
- You manage your own compute resources in the enviroment
- Resources like AWS ECS cluster
- Redirect batch to use your own managed resources
- Managed Compute Environment
- High or low priority environment