ECS Cluster Modes

Ref: https://learn.cantrill.io/courses/1820301/lectures/41301614

Amazon ECS - EC2 Mode

🔧 EC2 instances used as container hosts
- Containers deployed to instances (container hosts) via tasks & services
- ECS cluster created by specifying the number of EC2 instances with an ASG
  - ASG = AutoScaling Group (controls horizontal scaling for EC2 instances)
- ❗ Instances visible in EC2 (you can connect to them, stop them, etc)

Untitled

Great middle ground: ECS manages container overhead, customer manages container hosts
- 👍 EC2 pricing benefits can be applied → retains some flexibility
  - EC2 Reserved Instances (RIs) can be used as container hosts
  - EC2 Spot Instances can be used too
- 👎 Management overhead for container hosts (capacity & availability in the cluster)
  - ECS provisions the EC2 containers hosts, but customer expected to manage them
  - NOT serverless
  - ‼️ Once provisioned, you pay for the instances regardless of what containers consume!!
  - Container host management can be done through ECS tooling

AWS Fargate = Serverless Compute Engine → AWS service that offers a shared infrastructure platform to users
- ECS tasks & services can be allocated inside the AWS Fargate platform
- ❗ Even if using the same shared pool of HW resources, Fargate users are isolated
  - no visibility of other customers
  - 💡 similar to how EC2 users are isolated from each other when using shared hosts
🔧 With ECS-Fargate mode, container hosts are managed by AWS (NOT by customers)
- “Serverless” → no overhead for managing container hosts
  - No need to manage provisioning, capacity & availability in the cluster
- 👍 Customers only pay based on resources consumed by running tasks
  - no visibility of host costs

Untitled

ECS Cluster Mode	Container host location	Container host management	Billing
EC2	EC2 instances	Customer	Pay for whole instances (regardless of container usage)
Fargate	AWS Fargate shared infrastructure	AWS (Fargate service)	Pay based on resources consumed by running tasks