Backups that get tested, not just scheduled.
Meridian Labs runs simulation and data-processing jobs for academic partners. Their workloads are bursty, their datasets are large, and a lost week of results is a lost grant milestone.
Batch workers plus one coordinator.
From snapshot to serving traffic.
Locked snapshots for funded projects.
The problem
The lab's previous setup had backups, technically. They were weekly, stored on the same host, and had never been restored. When a disk failed during a deadline week, the most recent usable copy was six days old.
The move
We moved the batch fleet to nine Cloud VPS instances across two regions, each with daily snapshots and a 90-day retention window. Project data is written to a coordinator server with locked snapshots that cannot be deleted before the retention date, even by an admin.
Now the lab runs a restore drill at the start of every project: pick a snapshot, restore it to a new server, run a checksum against the source, destroy the copy. It takes about four minutes and it is part of the onboarding checklist.
The result
Two disk failures since the move, both recovered from the previous night's snapshot with no lost results. The drill is boring now, which is the point.
Sara M. · Operations lead, Meridian Labs
Stack
Talk to the team.
We will walk through your current setup and give you a straight answer about fit.