When should you replace your business server?

Age alone is a poor reason. Warranty status, failing disks, an unsupported operating system and a workload that has outgrown the hardware are much better ones.

Most servers we are called to look at are not broken. They are slow, out of warranty, running something Microsoft stopped supporting, and holding data the business cannot work without. The question is not whether the box still switches on.

Signs it is time

  • The operating system no longer receives security updates
  • The hardware is out of warranty, and spares would take days to source
  • Disks are reporting errors, or the RAID has been running degraded
  • Month-end or year-end reports have become noticeably slower each year
  • Memory is fully used and the system spends its time swapping
  • Nobody knows how long a rebuild would take if it failed tomorrow

Signs it is not, yet

  • It is simply old, but supported, healthy and comfortably fast enough.
  • The bottleneck is actually the network, or a workstation with a mechanical hard disk.
  • A memory or SSD upgrade would solve the specific complaint for a fraction of the cost.

We would rather tell you to add memory than sell you a server. It also builds the kind of relationship where you believe us when we do say it is time.

  1. AssessWorkload and health
  2. SpecifySized for real usage
  3. BuildInstall and configure
  4. MigratePlanned window
  5. SupportMonitored afterwards
A replacement is a short project, not just a purchase.

Size it from measurements, not habit

The useful inputs are: the number of concurrent users, the size of the application data and how quickly it grows, the CPU and memory actually in use at the busiest time of day, and whether the workload is limited by storage speed. A machine specified from those numbers usually costs less and performs better than one specified from the old server's sticker.

One specification decision is worth spending time on: storage. Most business applications on small servers are limited by how quickly storage responds, not by processor speed. Moving from mechanical disks to SSDs is usually the single change people notice on day one — more than doubling the core count.

Virtualize while you are there

If you are buying a new server anyway, it costs nothing extra to install a hypervisor and run the workloads as virtual machines. The benefit arrives at the next replacement: instead of reinstalling everything, the machines are copied to the new host. It also means a whole server can be backed up and restored as one unit, which shortens recovery from days to hours.

Plan the move before you order

  1. Agree a windowMost migrations happen in the evening or over a weekend.
  2. Copy and verify firstData moves ahead of time; only the final differences move during the window.
  3. Keep the old serverPowered off but intact for a few weeks, as a fallback.
  4. Check the small thingsPrinters, scanners, licences, scheduled tasks and backup jobs all need to follow.

Common mistakes

  • Buying a desktop-class machine for a server role because the price looked better
  • Reusing the old disks in the new server
  • Forgetting the backup job, so the new server runs unprotected for weeks
  • Decommissioning the old machine on the same day as the cut-over

More IT guides

Want this reviewed for your business?

We'll look at what you have, explain the practical options and tell you what we would do.