Home
About us
Solutions
Case studies
News
Case study · Industry
From on-premise to ROSA
Five years of cloud journey, start to finish:
the OpenShift environment moves to AWS.
Industry · from on-premise OpenShift to Red Hat OpenShift Service on AWS · about three months
Diagram of the work. The project reports about three months, from the technical analysis to the end of the migration.
The need
The last step
is the most delicate.
The client began a cloud migration with us that lasted about five years, with the goal of making its systems cloud native. We were there for every stage, from the first containerisation to this final one: moving the on-premise OpenShift environment onto ROSA (Red Hat OpenShift Service on AWS), while dealing with scalability, operational continuity, infrastructure efficiency and the modernisation of the release process.
Scalability
The infrastructure had to follow the load, instead of being sized once and for all.
Operational continuity
The move could not interrupt the live services or lose the connections to the legacy systems.
Infrastructure efficiency
Consumption and peaks have to be governed: in the cloud every sizing decision has an immediate cost.
Releases to modernise
The move was also the chance to make deployments repeatable and automated.
What we did
Size, migrate,
connect, automate.
About three months from the technical analysis to completion, with the trickiest part — network, Integration Layer and Single Sign-On — tackled head on.
The project in full
Three months,
from analysis to cutting the ribbon.
Add Value delivered the migration in about three months, starting from the technical and sizing analysis of the infrastructure and from defining the AWS resources needed. The applications were migrated, scripts and pipelines built with Helm and Argo, and the network and connectivity questions between the ROSA environment and the legacy resources dealt with.
Particular attention went to the Integration Layer, the Single Sign-On system and the adoption of Infrastructure as Code. Hybrid architectures were designed, the release pipelines automated, and health checks introduced that act by themselves when critical components saturate.
Before and after
Same applications,
different infrastructure.
Drag the handle: on the left the on-premise OpenShift cluster, on the right the ROSA environment on AWS, with automated release pipelines and health checks on the critical components.
- Length
- about 3 months
- Platform
- ROSA — OpenShift on AWS
- Releases
- Helm and Argo pipelines
- Infrastructure
- described as code
Schematic representation of the two environments, not a real topology.
The result
Infrastructure that
rewrites itself.
In three months, from analysis to completion, the infrastructure moved onto AWS and Red Hat OpenShift. With Infrastructure as Code every environment is described in versioned code: it can be recreated from scratch when needed, instead of being patched by hand. The health checks act on saturation by themselves, before it turns into an outage.
Releases automated with Helm and Argo, repeatable environment by environment
Infrastructure as Code: the environment is recreated from versioned code
Automatic health checks on consumption peaks, with no manual intervention
New infrastructure skills established in the team
A base ready for the next step: migrating the ERP to the cloud
The project also let the team build new infrastructure skills, and is the base for the next stage of the company’s cloud journey: moving the ERP from on-premise to cloud.
Inside the project
Three ways
of looking at it.
Whoever keeps the infrastructure up, whoever releases, whoever answers for continuity: one migration, three priorities.
Sized on the real numbers
The work started from the technical and sizing analysis of the existing infrastructure and from defining the AWS resources needed: sizing before the migration, not after the first bill.
- Technical and sizing analysis of what was there
- Hybrid architectures designed, not inherited
- Infrastructure as Code as the default
Pipelines instead of procedures
Scripts and pipelines built with Helm and Argo, provisioning and deploy automated: a release stops being an event and becomes a repeatable operation.
- Automated release pipelines
- Helm and Argo for scripts and deploy
- The same release in every environment
It notices on its own
Health checks were introduced that act automatically when critical components saturate: continuity no longer depends on who happens to be watching the dashboard.
- Health checks that act automatically
- Consumption and spikes handled
- Connectivity to the legacy resources guaranteed
Technologies and areas
Keep reading
Other stories
Banking · Software
Corporate Remote Banking
A multi-bank, multi-company platform with a modern interface and a microservices back-end, built for a technology provider to financial institutions.
0 outages
The full caseBanking · Data
Data migration machine
Automates diagnostics, normalisation and transfer of the data needed to migrate NPL portfolios and support later acquisitions.
~350 transformations · 10 verticals
The full caseBanking · Software
NPL field agents app
The Android app and the web portal, whose features partly overlapped, were brought together into one Angular Progressive Web App for the field agents and the recovery firms.
from 2 applications to 1
The full caseInsurance · Software
Regression testing
An automated testing service running since 2019, first with UFT and then with Playwright, to check that new features do not break the existing ones.
+3,700 automated tests
The full caseBanking · Software
Exchange Digital Corporate
The Corporate Banking flow backbone was rebuilt and run in parallel with the old system, handling around 1.5 million flows a day.
from 10 minutes to under 1
The full caseInsurance · Software
API Management middleware
A custom .NET API Manager with a two-way orchestrator and ELK monitoring, for near real-time policy issuing, 24/7.
+20% performance
The full caseInsurance · Software
Life AM
Running and maintaining the life insurance software: corrective and evolutionary work, plus the one-off requests.
97% SLAs met
The full case