Success Story
Azure-to-AWS migration in 7 months and continuous FinOps: ~50% fewer EKS instances
Alephee is an e-commerce software company founded in 2017 in Buenos Aires. Its platform connects brands, manufacturers and resellers in the automotive aftermarket (clients like General Motors and Bosch) with LatAm marketplaces such as Mercado Libre and Amazon, centralizing catalog, stock, pricing, orders and real-time reporting. High sales volume and B2B/B2C integrations demand total availability.
Watch the story
7 months
Azure→AWS migration with no data loss
~50%
fewer EKS instances (Karpenter + Spot)
3+ years
relationship (migration + FinOps + MSP)
Challenge
Alephee ran on a fragmented, costly and hard-to-scale infrastructure on Azure, with scattered technologies (AKS, Docker Swarm, App Services, SQL Server VMs) that created high maintenance overhead. It needed to modernize toward a cloud-native model on AWS without losing availability. After the migration, new challenges appeared: high costs from idle and oversized resources, and dependency on legacy architectures (x86, .NET nearing end of life).
Solution
Craftech migrated the entire platform from Azure to AWS in an orderly, uninterrupted way (AKS and Docker Swarm to Amazon EKS; App Services to Elastic Beanstalk with Blue/Green; SQL Server to Amazon RDS via AWS DMS; Blob Storage to S3), all with Terraform and adapted CI/CD. It then stayed as the continuous operations team (MSP + DevOps), with an aggressive FinOps strategy: Karpenter + Spot, automatic nightly shutdown of DEV/Test, rightsizing with Goldilocks and Kubecost, and IOPS tuning on MongoDB Atlas. Today it's modernizing toward Graviton (ARM64) and .NET 8 and exploring data analytics.
Results
The migration succeeded with no data loss and no major disruptions, with a consolidated AWS architecture and 100% integrated CI/CD. On FinOps: savings of ~US$850/mo in DEV and ~US$565/mo in PROD in the first cleanup phases, ~50% fewer EKS instances with Karpenter and +US$720/mo more from MongoDB Atlas IOPS tuning. Modernization left multi-architecture CI/CD pipelines, heavy workloads running on Graviton (better price/performance) and 30+ days of e-commerce metric retention without compromising cost.
Stack & technologies
