Success Story
Digiventures migrates to multi-account Amazon EKS and projects savings of around 40%
At a glance
Digiventures provides digital onboarding and lead capture solutions to financial institutions. Its architecture, in a single root account and outside AWS best practices, created operational and financial risks. In 4 months, Craftech migrated its microservices to Amazon EKS in a multi-account architecture, with projected savings of around 40% and faster deployments.
~40%
projected savings
1–2 min
API deployment (previously 5–6 min)
3–4 min
frontend deployment (previously 6–7 min)
Digiventures provides digital onboarding and lead capture solutions to financial institutions, such as banks and factoring companies, to digitize customer acquisition.
Solution implemented
- Designed a multi-account and EKS architecture following the AWS Well-Architected Framework, with segmented PRD, STG and QA environments
- Migrated the microservices from Elastic Beanstalk, Lambdas and EC2 to Amazon EKS
- Implemented private networks with VPN-only access and no unnecessary public exposure
- Implemented GitOps with ArgoCD for more control, visibility and fast rollbacks
- Added observability with Grafana, Prometheus and Loki
- Defined most resources as code, ready for replication and DRP plans
Business Challenge
Digiventures' architecture fell outside AWS best practices, which created operational and financial risks. Everything was in a single root account, with QA, staging and production exposed to cross-impact, uncontrolled access and little traceability. Critical resources were unnecessarily exposed to the internet. Elastic Beanstalk and EC2 meant high monthly spend and slow deployments, of 5 to 7 minutes, with little visibility. There were also limits on scaling horizontally.
Solution
In 4 months, Craftech migrated the microservices (from Elastic Beanstalk, Lambdas and EC2) to a multi-environment, multi-account architecture on Amazon EKS, designed following the AWS Well-Architected Framework, with segmented PRD, STG and QA environments. Resources were placed in private networks with VPN-only access and no unnecessary public exposure. It implemented GitOps with ArgoCD for more control, visibility and fast rollbacks, and observability with Grafana, Prometheus and Loki. Most resources were defined as code, ready for replication and DRP plans.
Outcomes
- Projected savings of around 40%, with ROI projected in 4 months once the previous infrastructure is decommissioned
- API deployment down from 5–6 minutes to 1–2 minutes and frontend deployment from 6–7 minutes to 3–4 minutes
- No risk of cross-impact or public exposure of services: better security and governance
- A platform on Amazon EKS ready to scale and for future optimizations, such as ARM instances and shutdown policies
Stack & technologies
