AWS Container Services: ECS, EKS, and Fargate

This study guide covers the core container orchestration and compute services provided by AWS, focusing on architectural decision-making for modernizing applications as microservices.

Learning Objectives

• Differentiate between Amazon ECS and Amazon EKS orchestration models.
• Identify use cases for AWS Fargate versus EC2 launch types.
• Determine the appropriate monitoring solution for containerized workloads.
• Explain the shared responsibility model regarding the control plane and data plane.

Key Terms & Glossary

• Orchestration: The automated management of container lifecycles, including deployment, scaling, and networking.
• Control Plane: The "brains" of the service that manages the cluster state, scheduling, and API (managed by AWS in ECS/EKS).
• Data Plane: The actual compute resources (EC2 instances or Fargate) where containers run.
• Task (ECS): A grouping of one or more containers that run together on the same host.
• Pod (EKS): The smallest deployable units of computing that you can create and manage in Kubernetes.
• Serverless (Fargate): A compute engine for containers that allows you to run containers without managing servers or clusters.

The "Big Idea"

Containerization is the primary vehicle for application modernization. By decoupling a monolith into microservices, you gain agility and scalability. AWS offers two "flavors" of orchestration: ECS (AWS-Native simplicity) and EKS (Open-source Kubernetes flexibility). Regardless of the orchestrator, you choose how to power them: either by managing your own EC2 instances or using the Fargate serverless engine to remove operational overhead.

Formula / Concept Box

Hierarchical Outline

• I. Amazon Elastic Container Service (ECS)
  • AWS Native: Deep integration with IAM, ELB, and CloudWatch.
  • Opinionated: Simplifies the container journey with a lower learning curve.
  • Windows Support: Managed AMIs for Windows and Linux.
• II. Amazon Elastic Kubernetes Service (EKS)
  • Vanilla Kubernetes: Supports standard K8s tools and open-source ecosystems.
  • Hybrid/Multi-cloud: Easier to migrate workloads from on-premises or other clouds.
  • Complexity: Requires Kubernetes-specific expertise.
• III. Compute Platforms (Data Plane)
  • EC2 Launch Type: Full control over patching, instance types, and overcommitting resources.
  • Fargate Launch Type: Serverless, no OS management, but lacks persistent local storage (ephemeral).

Visual Anchors

Container Service Decision Flow

Control Plane vs. Data Plane Abstraction

Definition-Example Pairs

• Service Integration: The ability of a container service to interact natively with other AWS tools.
  • Example: Using an IAM Role for Tasks in ECS to allow a container to write specifically to an S3 bucket without using long-term credentials.
• Ephemeral Storage: Temporary storage that exists only for the lifetime of the container.
  • Example: A Fargate task processing a temporary log file. If the task restarts, the log file is deleted unless saved to EFS.
• Container Insights: A CloudWatch feature providing granular metrics.
  • Example: Diagnosing a "Pod flapping" issue in EKS by viewing restart failure reasons in the CloudWatch dashboard.

Worked Examples

Scenario 1: The Migration Specialist

Problem: A company is migrating a legacy Java application. They have no experience with Kubernetes and want to integrate with their existing AWS infrastructure with minimal friction. Solution: Recommend Amazon ECS. Because it is "opinionated" and native to AWS, the learning curve is shorter, and it integrates seamlessly with ALB and IAM.

Scenario 2: The Serverless Strategy

Problem: A startup wants to run a containerized API but has no DevOps staff to patch Linux kernels or manage EC2 scaling groups. Solution: Deploy on ECS or EKS using the Fargate Launch Type. Fargate removes the operational overhead of managing the host OS and infrastructure lifecycle.

Checkpoint Questions

1. Which container service is best if you require "vanilla" Kubernetes for portability?
2. What is the main risk of using Fargate for stateful applications that don't use external storage?
3. True or False: Amazon ECS supports Windows containers.
4. Which monitoring tool provides container-level visibility like task/pod CPU and memory utilization?

[!TIP] Answers: 1. Amazon EKS; 2. Data loss (storage is ephemeral); 3. True; 4. AWS CloudWatch Container Insights.

Muddy Points & Cross-Refs

• Fargate vs. Lambda: Both are serverless, but Fargate is for standard Docker containers (long-running or complex), while Lambda is for event-driven functions.
• Persistence on Fargate: To save data, you must mount external storage like Amazon EFS. Local disk is wiped on termination.
• Learning Curve: EKS is powerful but complex. For the SAP-C02 exam, remember: EKS = Flexibility/K8s, ECS = Simplicity/Native.

Comparison Tables

ECS vs. EKS

EC2 Launch Type vs. Fargate