My First AWS Project — Scalable Infrastructure for a Real-Time Auction Platform

Lead Operations Engineer – DevOps · Network Redux

My first AWS project — and the first customer onboarded to Network Redux’s Managed Cloud Services. Built a scalable infrastructure for a US-based online auction marketplace and real-time bidding platform. What started as a classic AWS stack in 2016 evolved over the years into a fully modernized, containerized platform on Amazon EKS.

Context

The customer needed a reliable, auto-scaling infrastructure to handle unpredictable traffic spikes during live auctions. Real-time bidding demanded low-latency database operations and the ability to scale compute on demand. This was 2015 — many AWS features we take for granted today didn’t exist yet.

What I Did

Initial Build (2015)

  • Designed the full AWS architecture — Route 53, ELB, EC2 Auto Scaling Groups, RDS, and DynamoDB
  • Configured Amazon DynamoDB for the real-time auction engine
  • Set up 3rd-party CDN for image caching and static asset delivery
  • Built Ansible playbooks for baking AMIs with Nginx and PHP stacks
  • Implemented CI/CD with AWS CodeDeploy and CodePipeline
  • Provisioned all resources using CloudFormation templates — reverse-engineered from a YAML file left by a previous contractor on a bastion host, then rewritten and standardized. Those templates became the foundation for most stacks we provisioned at Network Redux

The DynamoDB Challenge

  • DynamoDB auto-scaling didn’t exist at the time. Implemented Dynamic DynamoDB — a script-based solution that monitored DynamoDB metrics and scaled tables up/down based on custom policies. It worked like a charm.

Evolution Over the Years

  • This customer was always the first to green-light upgrades whenever AWS introduced new services
  • Over time, migrated the entire platform to a modernized, containerized architecture:
    • Amazon EKS for container orchestration
    • Karpenter with Spot instance fleet for cost-efficient compute
    • Aurora Serverless replacing RDS
    • ArgoCD for GitOps-based deployments

Results

MetricDetail
Starting pointClassic EC2 + RDS + DynamoDB stack (2015)
Final stateFully containerized on EKS with Karpenter + ArgoCD
DynamoDB scalingCustom auto-scaling before AWS native support existed
Infrastructure as CodeCloudFormation templates that became the standard for Network Redux
Customer relationshipMulti-year engagement — continuously modernized

A happy, forward-thinking customer that helped me kick-start solution engineering on AWS Cloud. Always grateful to Randy Ehli.

Stack

2015: Route 53 · ELB · EC2 Auto Scaling · DynamoDB · RDS · CloudFormation · Ansible · CodeDeploy · CodePipeline · Dynamic DynamoDB

Final: Amazon EKS · Karpenter · Spot Instances · Aurora Serverless · ArgoCD · GitOps