Project
Multi-datacenter Kubernetes platform
On-premises Kubernetes spread across independent datacenters, with workload and security domains separated by design.
What it is
A Kubernetes platform for telecommunications workloads, spread across independent on-premises datacenters. I contribute to the architecture within a collaborative team and work on the platform day to day.
Why it exists
The workloads need to run close to existing telecom systems and can’t simply be handed to a managed public-cloud Kubernetes service. That pushes the interesting problems — networking, storage, failure domains, operations — back into the platform team.
How it works
- Workloads are grouped into distinct clusters by purpose and security boundary: supporting services, backend services, and externally-facing applications.
- Cilium provides cluster networking, with Hubble for visibility.
- Each datacenter keeps its own storage architecture so data stays local where it needs to be.
- Internal service domains and externally reachable application domains are kept separate for DNS, ingress and load balancing.
- Independent datacenter failure domains allow for inter-datacenter failover.
Notes
- On Proxmox, I manage the VMs the platform runs on and contribute to architectural discussions — I don’t administer the hypervisor platform itself.
- The design trade-offs are the topic of the draft write-up.
Status
Active. Architecture and ongoing engineering across multiple datacenters.