Why Most Small Offices Are Running Too Much Hardware
The pattern I walk into repeatedly looks like this. There is a domain controller on a tower from 2016. There is a second box that exists only because a line-of-business application shipped with an installer that demanded its own machine. There is a third that runs a backup agent and nothing else. Each one has its own power draw, its own failing fan, its own out-of-warranty support contract, and its own separate way of dying at an inconvenient hour.
Every one of those machines is idling. Every small-office file or application server I have measured idles in single-digit CPU and well under half its installed RAM. You are paying for capacity three times over and using it once. Virtualization is the fix that has been standard practice in enterprise for twenty years and still has not fully reached the ten-to-fifty-seat office, mostly because nobody has walked the owner through what the migration actually involves.
What I do is collapse those separate boxes into virtual machines running on one properly specified host, or on a small cluster if uptime matters enough to justify it. The workloads stay exactly where the staff expect them. The same server names, the same shares, the same printers. What changes is what sits underneath, and what happens when a disk fails.
This page goes deeper on the hypervisor side specifically. If you want the wider picture of hardware specification, hosting panels, firewalls and NAS work, that lives on server builds and infrastructure.
Proxmox, and Getting Off VMware
I standardised on Proxmox and I no longer run VMware. Since the Broadcom acquisition changed the licensing, the question I actually get asked is not which hypervisor is better, it is what it costs to leave.
Proxmox VE
Open source, Debian based, no per-socket license to renew. You get the full feature set on day one: live migration, ZFS with checksummed storage and instant snapshots, LXC containers alongside full VMs, and a clustered web interface that does not need a separate management appliance. Proxmox Backup Server does deduplicated incremental backups with verifiable restores, and it is free to run.
This is what I recommend for most small and mid-size offices. It is also what I run for my own production hosting, so I am not selling something I do not use.
Migrating off VMware
If you are on ESXi and your renewal quote arrived with a number nobody budgeted for, that is the common case right now. Broadcom moved VMware to subscription bundles and priced the small estate out, and a ten-to-fifty-seat office is exactly the size that gets hit hardest.
I do not run VMware and I am not going to sell you a renewal. What I do is the move: export the VMs, bring them up on Proxmox, sort the guest drivers and the boot mode, and check what your application vendors will actually support before anything is committed. If the honest answer is that you should stay put for another cycle, I will tell you that instead.
Moving Physical Servers to Virtual
Physical-to-virtual conversion is the part owners worry about, and the worry is reasonable. The old machine is holding the business. Here is how I run it so the risk stays low.
Inventory and Sizing
I document every service, share, scheduled task, license key, static IP and vendor dependency on each existing box before anything is touched. Then I size the host from measured load, not from adding up the spec sheets of the machines being replaced. Under-sizing is the mistake that makes people distrust virtualization.
Convert and Test in Parallel
The conversion runs while the physical server keeps working. The resulting VM boots on an isolated network segment where I fix boot drivers, install virtio or VMware tools, correct storage controller settings, and let you log in and test the actual application. Nothing is cut over until the copy has been proven.
Cutover With a Way Back
A short scheduled window, usually an evening, to take a final delta sync and move the IP and name across. The original hardware is powered down, not wiped, and stays on the shelf for a defined period. If something surfaces on Monday morning that testing missed, the rollback is plugging the old box back in.
Windows Server workloads carry extra care: activation, domain controller roles that should be rebuilt fresh rather than converted, and print servers that need their drivers reinstalled rather than carried over. Those specifics are covered on Windows Server management.
Snapshots Are Not Backups
This is the single most common misunderstanding I correct, and it is the one that costs businesses their data. A snapshot and a backup solve different problems and you need both.
What a Snapshot Is For
A point-in-time marker on the same storage as the VM. Perfect for a fifteen-minute undo before a patch, a driver install, or an application upgrade. Take it, do the risky thing, roll back or drop it. It costs seconds.
Why It Is Not a Backup
It lives on the same array as the thing it protects. A failed pool, a ransomware encryption event or a stolen host takes the snapshot with it. Long-lived snapshots also grow without limit and degrade VM performance until the disk fills.
The Backup Layer
Scheduled full-VM backups to separate storage, then a copy off site. I use Proxmox Backup Server for deduplicated incrementals, or NAS targets with retention tiers. Three copies, two media, one off site is still the rule that survives contact with reality.
Restore Testing
A backup nobody has restored is a rumour. I schedule verification jobs and periodically boot a restored VM in an isolated network to confirm it comes up and the application opens. That is the only proof that counts.
Where the Money Comes Back
Consolidation cuts spend in several places at once. One host to buy instead of three. One set of drives, one warranty, one support contract. Less rack space and less power and cooling, which shows up on a bill every month rather than once at purchase. Windows Server licensing also changes shape once you are virtual, and Datacenter edition can work out cheaper than several Standard licenses once the VM count climbs.
The savings that are harder to put on a spreadsheet are usually the larger ones. Rebuilding a failed physical server means sourcing hardware, reinstalling an operating system and restoring data, which is days. Restoring a VM to a different host is hours, and it does not care whether the replacement hardware matches. A new server that used to mean a purchase order and a lead time now means ten minutes of provisioning on capacity you already own.
I put real numbers against your actual estate rather than guessing at it. That exercise is what IT audit and cost reduction covers, and virtualization is usually the biggest single line item that comes out of it.
Proof of Work
Production Infrastructure
The hypervisors, VPS fleet, reverse proxies and backup targets I run day to day. This is the estate the practices on this page were built against, not a slide deck.
Server Builds
Host specification, hosting panels, firewall rules, storage layout and NAS administration. The layer the hypervisor sits on, and the wider service this page drills into.
Self-Hosted Gitea Server
A working example of a service that used to justify its own machine and now lives in a container on shared capacity, with backups and a documented restore path.
Ready to Consolidate Your Servers?
Tell me what you are running now and what keeps breaking. A 24 hour answering service picks up and routes straight to my cell when something is down; scheduled work happens in agreed windows. I am based in Lynchburg, onsite across Central Virginia, and remote for clients nationwide.
Get in Touch