The 5-year lifecycle of our virtualization platform

In 2018, we decided to manage our virtualization infrastructure fully in-house. Instead of relying on external providers, we built and maintained everything within our own datacenters in Luxembourg. This gave us full control over reliability, performance and security. It also allowed us to follow a clear technical direction rooted in open-source principles.

We treat infrastructure as a long-term investment. Every five years, we evaluate the lifecycle of our hardware and software stack: what still works, what doesn’t and what new technologies are worth adopting. This approach helps us stay stable while continuing to evolve.

2018 – Starting simple: KVM and OpenNebula

Our first platform was based on five physical servers. We ran KVM as our hypervisor and OpenNebula for orchestration. It was efficient, lightweight and well suited for the initial scale of our internal services and customer workloads.

But as our infrastructure expanded, we needed better failover mechanisms, improved storage performance and more flexibility in how we managed capacity.

2019 – Scaling with OpenStack

To meet growing demand, we deployed a new platform:

20 servers across two datacenters
Redundant storage arrays
Upgraded internal networking
Migration to OpenStack for orchestration
Capacity scaled to over 500 virtual machines

OpenStack gave us the scalability and control we needed. The transition required precise planning and configuration, but our team ensured a smooth deployment. This setup served us well for several years and laid the groundwork for operating geo-redundant infrastructure.

2021 – Major upgrade, smoother operations

Two years later, we upgraded to a newer version of OpenStack. We built a test cluster using two servers per datacenter and migrated over 200 virtual machines across six months carefully and with nearly zero downtime.

This allowed us to:

Enable hot-add for CPU and RAM
Improve VM mobility and live migration
Modernize our orchestration layer without disrupting service

This upgrade reinforced a key principle: deep testing and methodical rollouts are essential when dealing with critical infrastructure.

2024 – New generation with Proxmox and CEPH

In 2024, we began a new lifecycle. Some hardware had reached end-of-life, and shifts in the virtualization landscape including Broadcom’s acquisition of VMware confirmed our decision to avoid overcomplicated or vendor-locked solutions.

We adopted:

Proxmox VE for virtualization and orchestration
CEPH for distributed, redundant storage
6 Proxmox servers and 8 CEPH storage nodes across two sites

The result: a cleaner architecture, more efficient hardware usage and simpler management. All while improving performance and fault tolerance.

What Proxmox and CEPH bring

These technologies fit our environment not just technically, but operationally. They are stable, open-source and well supported with active communities and clear roadmaps.

Proxmox VE

Built-in clustering and automatic failover
Web UI with full CLI/API access
Reliable live migration and role-based access control

CEPH

Self-healing, distributed storage
Native support for geo-redundancy
Scales horizontally without rework or downtime

Together, they give us a platform that’s easier to maintain and more resilient by design reducing manual overhead and complexity, even as we grow.

What drives our decisions

We don’t follow trends for the sake of it but we stay alert to what the industry offers. Every decision we make is backed by testing, operational experience and long-term thinking.

We favor technologies we can operate confidently
We build systems our team can maintain over years, not quarters
We balance innovation with reliability and control
We continuously monitor, review and adjust where needed
And we stay true to our open-source mindset, because it works

Our latest platform is the result of five years of lessons learned, tested improvements and consistent engineering. It’s not just an upgrade. It’s a step forward that gives us and our customers a stronger foundation for what’s next.