Skip to main content

Amazon ECS Costs: Service Scaling, Fargate Sizing and Spot

Amazon ECS orchestration carries no charge. Every dollar on an ECS bill is capacity: EC2 instances, or Fargate vCPU, memory and storage.

Where ECS spend comes from

AWS states there is no additional charge for ECS orchestration. With the EC2 launch type you pay for the instances, volumes and public IPv4 addresses you run. With Fargate you pay for the vCPU, memory and storage your tasks request, with a one-minute minimum. So the two levers are how many tasks run, and how large each task is.

Scaling a service down on a timer

ECS is on ZopNight’s Start/Stop supported list as service scaling. A stop scales the service’s desired task count down and a start restores it. ZopNight discovers ECS clusters and services as a parent-child hierarchy and detects whether each service runs on Fargate or EC2 capacity providers, so a schedule can be attached to one service or to a group of them. With Fargate, fewer running tasks is directly fewer billed vCPU and memory seconds. With the EC2 launch type the saving only lands once the instances underneath also scale in, so pair the service schedule with the cluster’s Auto Scaling group.

ZopNight can also own the scaling policy. Its autoscaler manages ECS services through Application Auto Scaling, with policies prefixed so they can be removed cleanly and the previous configuration restored. If a service already has scaling policies, you choose whether ZopNight adopts them unchanged or replaces them.

ZopNight’s ECS rules

  • ECS Fargate Oversized Task prices shrinking the task’s vCPU term using the real per-task rate difference.
  • ECS Service CPU Underutilized does the same from 30 days of CPU and memory.
  • Running ECS Service Idle (No Traffic) requires both zero requests and idle compute before it fires.
  • ECS Fargate Spot Opportunity for Dev/Test uses live Spot and on-demand rates, never a flat discount.
  • ECS Service Auto Scaling Not Enabled, Empty ECS Cluster and ECS Cluster Missing Container Insights cover configuration.

Planned peaks

For a sale or launch, event readiness can pre-scale an ECS service using Application Auto Scaling scheduled actions and restore the original minimum, maximum and desired values afterwards.

Fact sheets: ECS service and ECS cluster. Scheduling notes: ECS service schedules. Pre-scaling is covered on event readiness.

Stop watching the waste.
Start cutting it.

See. Find. Fix. Automatic.

Connect your first cloud account in under 5 minutes. See your first remediation in under 7. No credit card required.

CDCR connect detect classify remediate
full audit every action traceable
read-only default access
Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console· Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console·