Operations
How we plan maintenance windows without surprises
The rules, the sign-off and what the numbers looked like last quarter.
Infrastructure
On 21 October we cut Vienna and Copenhagen over from rented edge capacity to routers we operate ourselves. Both locations have been carrying customer traffic since, and the two new racks in AMS2 are now live. Nothing about your configuration needs to change.
The short version:
The longer version involves a very warm riser, a fibre pull that took longer than the rack build, and a label printer that ran out at the least convenient moment. The invoice is less interesting than the lesson: never schedule a cutover on the same day as a hardware delivery.
AMS2 is already our European peering hub. Adding racks there gives us room for the compute expansion planned for 2027 without spending the next year apologising for capacity. The alternative was another rented suite in Frankfurt, which would have added latency and cost without adding resilience.
It also means we can terminate private network attachments in Amsterdam without hairpinning traffic through London. For customers running services in both cities, that is a small but real improvement in both directions.
If you have monitoring that alarms on traceroute changes, you may see the new path appear. That is expected; the destination did not move.
In the first quarter of 2027 we will add a second path into Manchester and upgrade the Dublin edge router. After that, the last rented edge circuit disappears from the estate. Each window will be posted on the status page at least five working days ahead, with a rollback plan and a named owner.
If you want the detailed latency data, it is in the monthly customer report. If you want to argue about our choice of router vendor, email Tom. He enjoys that.