Accelerated Cloud Migration for a Global Retailer

Northbridge's three colocation data centers, originally built in 2008, were running at 91% capacity and bleeding nearly $11M per year in hardware refresh, power, and circuit costs. The 2024 holiday season had narrowly avoided a catalog outage when their monolithic order-management system saturated its database connection pool at 04:12 EST on Black Friday morning -- the kind of near-miss that makes every CIO's stomach drop.

The leadership team gave engineering a non-negotiable mandate:

• Exit all three data centers before the next holiday season
• Cut total infrastructure spend by at least 30% within 18 months
• Achieve sub-200ms p95 checkout latency at 8x baseline traffic
• Zero unplanned downtime during the migration window

Complicating matters: 47 of the 247 in-scope applications had no current owner, 18 were running on unsupported Windows Server 2008 R2, and the core order-management platform was a 1.2 million LOC Java monolith with a single Oracle 11g backend that no one had patched in 26 months.

TekNinjas embedded a 14-person migration pod on-site in Dallas and ran a structured discovery using AWS Application Discovery Service plus a custom dependency mapper we wrote in Neo4j. Within six weeks we had a complete service graph and a wave-based migration plan grouping applications by blast radius and business criticality.

Migration approach

• Wave 1 (months 1-3): 88 stateless web tier apps rehosted via AWS MGN, validated with synthetic transactions in a parallel stack before cutover.
• Wave 2 (months 3-6): 102 internal back-office systems replatformed to ECS Fargate, with RDS Postgres replacing legacy MS SQL.
• Wave 3 (months 5-9): Core order-management strangled incrementally onto EKS using Spring Boot microservices behind an Istio service mesh, with Aurora PostgreSQL Global Database providing multi-region read replicas.

The Oracle data warehouse (8.4 TB) moved to Redshift via AWS DMS with a 12-hour cutover window over a Sunday, validated by row-count and checksum jobs we built into Airflow. Observability was unified on Datadog with custom SLO dashboards per application owner. CI/CD standardized on GitHub Actions with Terraform Cloud managing 100% of the new infrastructure.

Crucially, we kept Northbridge's internal teams in the driver's seat -- every wave included a four-week capability transfer with paired engineering and runbook authoring, so on-call ownership transferred at cutover.

The final wave cut over the weekend before Halloween 2025, four weeks ahead of the contractual milestone. The 2025 holiday season ran on the new platform with zero customer-impacting incidents.

• 247 applications migrated; 3 retired, 11 replatformed, 233 rehosted
• $4.7M in annual run-rate infrastructure savings (43% reduction)
• Black Friday p95 checkout latency dropped from 487ms to 178ms
• Mean deploy frequency rose from 1.4/week to 19/week across the engineering org
• Three colocation contracts terminated on schedule, freeing $2.1M in additional run-rate from year two onward