Deep Dives
The references you keep open while building.
Not blog posts. Long-form technical references — verified pricing, real math, copy-paste Terraform. Written for platform engineers who want the full picture, not a summary.
Deep Dive
AWS ECS Fargate: What It Is, How It Works, What It Costs
ECS schedules containers. Fargate runs them. Here's the full picture.
- →ECS schedules containers. Fargate runs them. You pay per task-second, not per instance.
- →On-demand: $0.04048/vCPU-hr + $0.004446/GB-hr (us-east-1). Fixed overhead per env: ~$50–150/mo.
- →At 10+ environments fixed costs dominate — ALB, NAT GW, CloudWatch compound per cluster.
Deep Dive
ECS Environment Scheduling: Stop Paying for Idle Fargate
Dev environments run 168 hrs/week. Your team works 40.
- →Dev environments run 168 hrs/week. Your team works 40. You're paying for 128 hrs of nothing.
- →3 native scheduling methods (EventBridge, Lambda, CLI) — all have the same problem at scale.
- →At 10+ environments you need fleet-level scheduling, not per-service cron jobs.
Deep Dive
AWS Fargate Spot: Pricing, Savings, and When to Actually Trust It
50–70% off on-demand. The part nobody explains is what happens when AWS takes it back.
- →Spot saves 50–70% vs on-demand. The price is variable — AWS adjusts based on supply/demand.
- →Default stopTimeout is 30s, not 120s. You lose 90s of your 2-min warning without knowing it.
- →ECS does not fall back to on-demand when Spot is unavailable. Tasks queue.
Deep Dive
ECS vs EKS: Which AWS Container Service?
The real gap isn't the $73/mo control plane. It's who patches nodes.
- →ECS control plane: $0. EKS: $73/mo per cluster. At 10 environments that's $730/mo before compute.
- →The real gap: who patches nodes, how fast containers start, whether you can stop dev to $0.
- →Choose ECS if you want containers without owning Kubernetes.
Deep Dive
ECS Fargate vs ECS EC2: The Break-Even Math
Same orchestrator, two compute models. Here's where each wins.
- →Same orchestrator, two compute models. Fargate: pay per task. EC2: pay per instance.
- →Fargate wins at <8 services or when you want to stop envs to $0. EC2 wins at 20+ always-on services.
- →The break-even math is honest — and fragile. One autoscaling misconfiguration flips it.
Deep Dive
ECS Fargate vs Kubernetes: The Honest Math
We ran Kubernetes. Here's why we stopped.
- →Kubernetes requires a platform team. ECS Fargate requires almost none.
- →EKS control plane + nodes + upgrades = ~0.5–2 FTE. ECS Fargate = ~0.05 FTE.
- →For AWS-only teams under 200 engineers, Fargate wins on cost and ops in almost every scenario.
Deep Dive
Migrating from AWS Proton
AWS Proton shuts down October 7, 2026. Here's what to do.
- →AWS Proton shuts down October 7, 2026. All Proton stacks need a replacement by then.
- →Fortem covers the ECS Fargate operations layer Proton provided — scheduling, cloning, visibility.
- →Migration timeline: 2–4 weeks depending on number of environments and templates.
See all content
Looking for shorter reads or comparisons?
Guides cover specific how-tos.
Versus pages cover why-not-something-else.
Use cases cover teams who already made the call.