Return to Project Vault

Software Architect & Full-Stack Developer | AHCC (2019)

Executive Summary

Re-architected and migrated an enterprise server infrastructure from legacy on-premise monolithic servers to a high-availability AWS cloud topology. Designed an asynchronous multi-phase database delta sync protocol to migrate a 550 GB relational database cluster with zero customer-facing downtime.

1. Context & Business Problem

  • ->Client / Domain: AHCC (Enterprise)
  • ->Timeline: 2019
  • ->Project Role: Software Architect & Full-Stack Developer

//The Problem

The client ran core business operations on single-point-of-failure on-premise servers with co-located relational databases and local disk storage dependencies. System backups caused severe performance drops, deployments required manual intervention, and scaling during traffic peaks was constrained by physical hardware limits.

2. Technical Stack & Infrastructure Architecture

  • ->Cloud Infrastructure: AWS ECS (containerized application tasks), S3 (media object storage), and CloudFront CDN.
  • ->Database & Caching: AWS Aurora Multi-AZ PostgreSQL 11 with PgBouncer connection pooling and Redis caching.
  • ->Automation & Provisioning: Declarative Terraform manifests for VPC networking, security policies, and container task definitions.

3. Architectural Decisions & Engineering Challenges

  • ->Multi-Phase Database Delta Sync: Configured asynchronous logical replication between on-premise PostgreSQL and cloud Aurora PostgreSQL, running row-level checksum verification scripts until replication lag stabilized below 15ms prior to DNS cutover.
  • ->Connection Spike Management: Placed PgBouncer in transaction pooling mode in front of the database cluster, allowing the system to handle bursts up to 4,500 concurrent connections without exhausting backend connection limits.
  • ->Split-Brain Prevention during Cutover: Lowered DNS TTL to 60 seconds 72 hours prior to migration and applied temporary read-only constraints on the source database at T-0 to reject stale write attempts during the final delta flush.
  • ->Stateless Application Decoupling: Migrated local media asset pipelines to S3 with pre-signed upload URLs and CloudFront distribution.

4. Operational & Institutional Impact

  • ->Zero Downtime Migration: Completed the production cutover within a scheduled sub-minute DNS switchover without data loss.
  • ->Database Latency: Reduced query response times substantially through Aurora indexing and read-replica distribution.
  • ->Automated Scalability: Reduced infrastructure provisioning time from days of manual setup to automated Terraform deployments.