How virtualization saved me and the company work
It was around 2007, and I was the IT Manager for the India operations of a HIPAA-compliant software company. The team designed and developed HIPAA-compliant applications across two offices in Chennai and several in the New York area around the HQ. I managed the Chennai side: Windows Server 2003 Active Directory — the PDC, the additional controllers — and the SQL Servers, with the two Chennai offices linked by a 1 Mbps local loop from Airtel and a 1 Mbps leased line to the internet through VSNL's higher tier. Two senior engineers, one per office, reported to me.
Two Dell servers, one power supply each
I was a physical server person at the time. Virtualization then meant something like VMware GSX running on top of an existing OS, and to me that looked like pure overhead — Xeon processors worked better talking directly to the hardware. We had two identical Dell servers, and for reasons long forgotten both had only a single power supply. No redundancy, no spare.
A Friday afternoon failure
On a Friday afternoon, a power event took out the domain controller's SMPS. That machine was the primary domain controller. If it stayed dead through the weekend, Monday morning nobody could authenticate, Visual Studio wouldn't work, and the team's output would be jeopardised. I raised the request with Dell straight away — but the part wasn't stocked locally. It had to come from Bangalore: two weeks.
Saturday at the office
So on Saturday, after failing the evening before, I was at the office. The other Dell was running the MS SQL database. I shut SQL down, pulled its power supply, moved it into the domain controller, and powered it on. Since the server had been cut off abruptly, it had to repair itself first — then it came up online. I tested user login and authentication, and it worked. What a relief.
The experiment that changed my mind
Relief wasn't a plan, though. The domain controller was now running on borrowed power and the SQL server had none. So I took a workstation and installed ESX on it — the bare-metal hypervisor this time, not GSX — and set up another desktop with the P2V converter on a Windows XP machine. I pointed it at the PDC, started the conversion with the recommended parameters, deployed it onto the workstation, and waited.
When the migration finished, I started the VM. The PDC booted. I shut the original PDC down immediately, enabled the VM's local network with a new static IP, fixed the minor repairs an abrupt shutdown always leaves behind, and tested user login and authentication again — it worked really well. The SQL server got its power supply back, booted like a charm, and the virtualized PDC took over for real.
That weekend flipped me. The domain controller now lived as something I could move, restart and rebuild — more resilient than the tin it used to run on, saved by hardware I'd previously dismissed as beneath it. From then on I stopped treating virtualization as overhead and started supporting it. Sometimes it takes one failed power supply to convert a physical-server person.