Professional experience at Tele2 Shared Services Center and Tele2 Sweden.
Problem
A legacy monitoring stack of Nagios components, cron scripts and hand-run deployments had to watch an ever-growing telecom network, and it couldn’t keep up with the volume of devices, metrics or users.
What I did
- Rebuilt collection. Replaced script-based SNMP polling with horizontally scalable Go polling services.
- Picked the right storage. Evaluated several long-term metrics stores and adopted ClickHouse for analytical workloads at this scale.
- Built a shared platform. A Kubernetes-based compute and data platform across three private data centres, with GitOps delivery, a shared Kafka cluster with per-team ACLs, and a discoverable RPC layer.
- Hybrid on purpose. Heavy metrics storage stayed on-prem for data-residency and cost reasons, with AWS used for processing and analytics where it made sense. That avoided an estimated 6–8× cost increase from moving everything to the cloud.
- Led the team. Lead engineer and mentor for five developers.
Results
From roughly 1 million devices and 200 users to 1–5 million devices and sensors and over 2,000 users, including sales, product and operations teams using new BI and analytics features. Release cadence moved from once or twice a month to several times a week.
Lessons
“Cloud vs on-prem” is the wrong question. The useful question is which workload belongs where, and why.