We saved a law firm's
20-year-old software
without touching it.

Three machines. Three single points of failure.

"We have a Windows 7 computer in the corner that runs Microsoft Money. Nobody's allowed to touch it. If it dies, we lose decades of financial history."

It wasn't just one PC. It was three.

• The "do not touch" PC — Windows 7, running Microsoft Money. No one remembered the admin password. Hard drive was 9 years old.
• The file server — aging Windows 10 box with every shared file, scan folder, and the TimeSlip billing database. No RAID. No offsite backup.
• The paralegal's workstation — WordPerfect X9 and Sage on a Windows 10 machine from 2014.

The firm ran on all three, every day. Replacement wasn't an option: Money hasn't been sold since 2009, WordPerfect licenses are non-transferable, and TimeSlip migration is a multi-month project.

The real risk: the Win7 drive was already showing reallocated sectors in CrystalDiskInfo. Not if it fails — when.

Freeze the computers in amber. Move them to safety.

We migrated the whole computer — OS, apps, registry, drivers, local files — into a virtual machine using Proxmox VE. Three times.

End state before we left the building:

Each PC runs as a locked-down kiosk VM. Users connect via RDP from their existing workstations. From their seat, nothing changed. We manage one host instead of three, with centralized backups and per-VM snapshots.

Infrastructure — before and after

How we ran the migration

We installed Tailscale and Win32-OpenSSH on each machine and ran a PowerShell audit script remotely: CPU, RAM, disk health via Get-PhysicalDisk, installed apps via Get-Package, active shares.

The locked Win7 box — no admin password on record — required the Utilman technique: boot recovery, swap Utilman.exe for cmd.exe, get a SYSTEM shell from the login screen, reset the password. 15 minutes.

# Audit script excerpt -- hardware inventory
Get-PhysicalDisk | Select FriendlyName,MediaType,HealthStatus,Size,OperationalStatus |
  ConvertTo-Json | Out-File "$env:TEMP\disk-health.json"

Get-WmiObject Win32_ComputerSystem |
  Select Name,TotalPhysicalMemory,NumberOfLogicalProcessors |
  ConvertTo-Json | Out-File "$env:TEMP\hw.json"

# Check VT-x / VT-d via msinfo32 export
msinfo32 /report "$env:TEMP\msinfo.txt" /categories +systemsummary

Disk2vhd (Sysinternals) captured each machine live via VSS — no shutdown, no disruption. Three .VHDX images totaling ~120 GB, staged to a portable drive over the network.

These images alone are a complete DR solution. Before the Proxmox host existed, all three machines were portable files restorable to any hardware in under an hour.

The healthiest of the three machines — i5-6500, 16 GB DDR4, Samsung 860 EVO — became the host. Pre-install prep:

• Verified VT-x and VT-d enabled in BIOS (required for full hardware virtualization and PCI passthrough)
• Cleaned chassis, replaced thermal paste, inspected PSU capacitors
• Confirmed SSD health via smartctl -a /dev/sda after Proxmox boot

# Proxmox VE 8.1 install -- USB boot, guided installer
# Target disk: /dev/sda (Samsung 860 EVO 500 GB)
# Filesystem: ext4 (ZFS not used -- single-disk host, no RAID)
# Management IP: 192.168.1.50/24 (static, assigned in BIOS DHCP reservation)

# Post-install: verify Intel VT-d is active
dmesg | grep -i "DMAR\|IOMMU"
# Expected: [    0.000000] DMAR: IOMMU enabled

New hardware purchased: $0.

Converted each .VHDX to .qcow2 using qemu-img on the host, then attached each disk to a new VM.

# Convert VHDX to qcow2 (run on Proxmox host after SCP of images)
qemu-img convert -f vhdx -O qcow2 /tmp/money-pc.vhdx \
  /var/lib/vz/images/100/vm-100-disk-0.qcow2

qemu-img convert -f vhdx -O qcow2 /tmp/server-pc.vhdx \
  /var/lib/vz/images/101/vm-101-disk-0.qcow2

qemu-img convert -f vhdx -O qcow2 /tmp/work-pc.vhdx \
  /var/lib/vz/images/102/vm-102-disk-0.qcow2

# VM config example (VM 100 = Money PC)
# /etc/pve/qemu-server/100.conf
# ---
# bios: seabios
# cores: 2
# memory: 4096
# net0: virtio=XX:XX:XX:XX:XX:XX,bridge=vmbr0
# scsi0: local:100/vm-100-disk-0.qcow2,size=40G
# scsihw: virtio-scsi-pci
# vga: std

VirtIO drivers injected into the Win7 VM post-boot via virtio-win ISO — significant disk and network performance improvement. Each VM received:

• Tailscale installed — persistent mesh node, auto-connect on boot
• RDP enabled — port 3389, Windows Firewall rule added
• Kiosk lockdown — Task Manager disabled via Group Policy, Start Menu restricted, auto-login configured, crash auto-restart enabled via sc failure
• Proxmox Backup Server job — nightly at 02:00, 7-day retention, all VMs

Each user got a labeled shortcut on their workstation desktop. mstsc.exe with pre-filled Tailscale IP.

# Shortcut target -- "Open Money" desktop icon
# C:\Windows\System32\mstsc.exe /v:100.x.x.x /f /w:1920 /h:1080
# 100.x.x.x = VM-Money's Tailscale IP
# /f = full-screen, /w /h = resolution match

# File shares remain accessible at original UNC paths:
# \\Server\Data     -> maps to VM-Server's C:\Data share
# \\Server\Scan     -> maps to VM-Server's C:\Scan share
# \\Server\TimeSlip -> maps to VM-Server's C:\TimeSlip share
# (Tailscale resolves "Server" via MagicDNS to VM-Server's node IP)

Double-click. App opens full-screen. Files where they always were. From their seat, nothing changed.

Before and after

Same software, same workflows, same files — on infrastructure we can replace without touching the apps inside. Host dies? Restore VMs to any $300 mini PC. Back online same day.

PCA migrated 3 machines in a saturday window. monday morning everyone signed in like nothing happened. we kept the proxmox host. the savings paid for the engagement in 5 months.

— managing partner, houston law firm (anonymized per engagement agreement)

Tools used

all tools listed ran on hardware the customer already owned. total software licensing cost for this engagement: $0.

Timeline

Why virtualization is the third option

A lot of small businesses run software that runs the company on hardware that's quietly dying. The usual options:

• "Just upgrade" — new version won't import old data, costs five figures, requires a year of retraining. For a 6-person firm, that's not a project. That's a crisis.
• "Just replace the PC" — install media gone, license non-transferable, vendor no longer exists.
• "Just live with it" — until the drive fails.

Virtualization is the third option. Move the entire machine — OS, apps, registry, drivers — into a container that runs on modern hardware and can be snapshotted like any other file. The app doesn't know anything changed. Hardware underneath gets swapped on a Saturday.

The PC becomes a file. The file is forever.