Servers and storage

The server side is calculated from the load: daily data volume, archive retention depth, acceptable downtime and recovery time after a failure. Everything else follows from those figures — the number and configuration of nodes, the type and capacity of storage, the backup design. The reverse order, where a model is picked first and the task fitted to it afterwards, ends either in running out of headroom within a year or in overpaying from day one.

Fault tolerance is designed separately: redundant power and cooling, backup links, a procedure for replacing disks without stopping the service. For a situation room or a dispatch centre this is not optional — the video wall shows what the server room feeds it, and its downtime is downtime for the whole hall.