Full health check
Once a year: systematic health check with a written report and prioritised actions. The minimum sensible programme for a standard enterprise server.
An hour of unplanned downtime typically costs far more than a day of planned maintenance — not because of what the maintenance costs, but because of what the unavailability costs. Preventive maintenance consolidates health checks, refresh, firmware updates and thermal optimisation into a coherent plan, calibrated on how critical the system is.
The realistic calculation, on enterprise systems: 1 hour of unplanned downtime on a critical server is worth, conservatively, the entire cost of a day of scheduled maintenance. On transactional or production systems it can be worth many times more. The question is not "whether" to run preventive maintenance, but "at what interval".
Once a year: systematic health check with a written report and prioritised actions. The minimum sensible programme for a standard enterprise server.
For systems under sustained heavy load, or datacenter environments with thermal conditions at the limit. Includes systematic thermal verification and fan curve calibration.
For systems 3+ years in production: a cycle combining scheduled hardware refresh, firmware updates against a compatibility matrix, and thermal optimisation. Typically every 18-24 months.
On this vertical (hardware) the standard model is per intervention with planned maintenance. For ongoing systems administration (Windows Server, Linux, VMware managed, monitoring, backup) the group's reference is AssistenzaServer.eu. For multi-client server estates we can define periodic health check programmes at a fixed interval.
Work on hot-swap components (drives, PSUs, some fans) is done without downtime. For work that requires taking the system down, a night or weekend window is agreed. Planning early makes it possible to handle both with no impact on applications.
It depends on the number of servers, how critical they are, the interval chosen, and whether on-site presence is required versus remote. On a single enterprise server the annual cost of basic preventive maintenance is a small fraction of the value of the system, and a smaller fraction still of the realistic cost of unplanned downtime.