Skip to content
Holits
Menu

Platform Engineering

Cutting cloud cost 40% on a European airline's booking platform

Moving NLB-fronted EC2 fleets to Auto Scaling groups behind ALBs, and sharing one environment between QA and GA with header-based routing.

Discuss a similar problem
By

The figures on this page are the engineer's own, from a contract engagement for a European airline's booking platform delivered through Amach. They are anonymised: the airline is not named.

Problem

The booking platform ran on EC2 fleets fronted by Network Load Balancers, with a full environment for each stage. That made environments expensive to run and slow to provision, and left release frequency and release reliability to improve.

What I did

  • Moved the NLB-fronted EC2 fleets to Auto Scaling groups behind Application Load Balancers.
  • Added header-based traffic steering at the ALB, so QA and GA can share one environment.
  • Wrote the infrastructure as Terragrunt modules.
  • Built the release path on Jenkins and GitHub Actions, with Packer images, blue/green and canary rollout.
  • Set up an OpenTelemetry stack for observability.

Result

  • Cloud cost down 40%, from sharing environments with header-based routing.
  • Provisioning from days to minutes.
  • Deployment frequency up 3x, with failed releases down 70%.

Stack

AWS (ALB, ASG, EC2), Terraform and Terragrunt, Jenkins, GitHub Actions, Packer, OpenTelemetry.