EC2 to ECS, at 12.5% of the cost

The company

CoinList is a cryptocurrency exchange — a platform where traffic is not smooth, and where being slow to scale during a spike is the same as being down.

The problem

Workloads ran on EC2 instances sized for peak traffic and left running through the troughs — the standard failure mode of lift-and-shift infrastructure. Exchange traffic is spiky by nature, so the gap between provisioned and used capacity was large and permanent.

Scaling the application took 45 minutes, which is long enough that autoscaling stops being a response to demand and becomes a guess about it.

What I did

I migrated the workloads from EC2 to containers on ECS, right-sizing resource allocation against actual usage rather than peak, and tuning autoscaling for the traffic spikes the platform genuinely sees.

I also led the CI/CD migration from Jenkins to GitHub Actions.

Result

• Web fleet cost down 87.5% — from $4/hour to $0.50/hour, measured by the client on one service’s web traffic during a low-traffic window.
• Application scaling time down 88% — from 45 minutes to about 5.

The second number matters more than the first. At 45 minutes, capacity decisions have to be made in advance and paid for continuously. At 5 minutes, they can be made in response to what is actually happening, which is what makes the cost reduction durable rather than a one-off trim.