I'd been talking about expanding beyond DC1 for years. In 2025, it finally happened. We're now running across three datacenter locations.
Why three locations? It's about options. Not being locked into one provider or location. Workloads can run active-active for high availability. Real geographic separation for disaster recovery. And off-site backups that are actually off-site.
Three datacenters
DC1: Worldstream Naaldwijk remains our primary datacenter. Most production workloads and primary database instances live here.
DC2: Databarn Amsterdam went live this year. 62 kilometers from DC1, enough distance for real geographic separation while staying close enough for low latency. We found the right partner to grow with in Oxar B.V. (And no, the underground bunker from our World Backup Day article was a joke.)
DC3: Colocenter Zoetermeer is our newest location. We racked this one right before Christmas. This is where the tape library will live for long-term archiving, and it serves as the backup site for DC1 and DC2. We're taking our time to get everything dialed in.
Under the hood
Three locations require serious monitoring and network infrastructure. This year we invested heavily in the things you don't normally see:
- Zabbix has been our monitoring backbone for years, but we spent significant time this year enhancing it. Better alerting, more metrics, and coverage across all three locations.
- LibreNMS gives us network visibility: switches, routers, bandwidth, and connectivity between sites.
- Mikrotik routers are being rolled out as dedicated routing infrastructure. More reliable than combining routing with other functions. We'll write more about the why and how of Mikrotik in a future article.
The 10G interconnect between DC1 and DC2 is nearly complete. We ran on Wireguard for a while, which worked fine, but a private interconnect brings latency under 2ms.
Ready for clients
Kubernetes we ran internally for years. In 2025, we deployed the first dedicated clusters for clients. In 2026, we're launching shared Kubernetes: more accessible, lower entry barrier, same reliability.
Barista, our control panel, took longer than expected. That's honest. But it's coming. In 2026, clients will be able to manage their own infrastructure and applications through Barista: resource insights, log access, and basic actions without having to contact us. That leaves more time for a cup of coffee.
2026
2025 was about strengthening the foundation: more resilience, more capacity, better tooling. 2026 will be about what you can see and use. With the infrastructure solid, we're shifting focus to customer-facing improvements:
- DC3 fully integrated
- Shared Kubernetes for clients
- Barista client access
- Experimenting with local AI models for monitoring and troubleshooting
Thanks to everyone who trusted us with their infrastructure this year. Here's to a solid 2026.