Scared to look in the box
Restores for legacy system often involve a day of reinstalling software, manually moving tapes around and sacrificing a goat to the storage gods so that the recovery will be completed. Legacy systems often do not have easily testable restores, leading to the "Schrödinger's back-up" problem: until data is lost, no-one really knows if it can be recovered.
Often the skills required to manage disaster recover and recovery of these legacy systems are lost, and the RPO and RTO become unknowns. In this case, the business is arbitraging the cost of embarrassing data loss and downtime with the relatively low cost of migrating the legacy machine to a shiny new virtual machine.
Perversely, in many environments the people championing retaining these old risky workloads are not the business management, but actual IT people trying to justify their lack of modern skill sets. As time goes on, the cost of migrating becomes greater as the skills required to migrate the obsolete tech become more rare.
As the risks become less understood, the true cost of operations increases until the box finally crashes.
Virtualisation changes the costs associated with management and agility for IT. Instead of needing KVMs or physical hands on a server to recover from a blue screen, even free hypervisors allow for quick out-of-band console access to troubleshoot issues.
Standardised virtual hardware abstractions mean that migrating from one generation of server to another requires just a few clicks, rather than a full rebuild and application-level migration. Storage array migrations that would have taken months now take hours or days. Abstracting the storage and management plane to the virtual level also frees us from vendor dependencies.
I can seamlessly move from one server, storage or network vendor to another without having to re-learn my day-to-day management tools and tasks. Testing of sketchy updates can be performed on clones of the virtual machine. This is important when I have an insanely flaky Tomcat server that carries a 20 per cent chance of successful update and an 80 per cent chance of blowing up and requiring a rebuild.
In the case of my recalcitrant Tomcat server, I can simply snapshot the virtual machine and make attempt after attempt until the upgrade goes through successfully. Easily cloneable cattle win out over pets quite clearly, here.
Virtualisation increases the time to "fail". Being able to "fail" quicker and cheaper means proof of concepts can happen quickly and keep the environment from growing old. The purpose of IT isn't to keep the lights on, it’s to enable the business to move quickly.
Legacy IT is a lethargic bottleneck on operations. There has to be a better reason for keeping dated IT around than preserving the jobs of the sacred priests maintaining the otherwise unknowable equipment.
Virtualisation is the first step. Automation and orchestration are the next. Self service for end users moves you into the cloud world and suddenly IT is delivering healthy services rather than tending to the needs and foibles of toxic legacy systems. ®