One of the initiatives we have underway in EMC IT is virtualizing our desktop infrastructure. And like most IT organizations, ours needs to be able to calculate a reasonable cost/benefit ratio before continuing beyond the pilot stage. Naturally, there are many consultant-provided cost and savings models available. Just as naturally, most focus on hardware equipment costs and savings. Then there’s the usual guideline that says operational costs are between double and triple the equipment costs.
But every IT shop has to verify that they either follow the rule, or are an exception. And that means lots of number crunching.
The good news is we have years of detailed data from help desk systems, and from detailed records kept by various functions within EMC IT. The Virtual Desktop Infrastructure (VDI) team within EMC IT is currently going over its penultimate set of figures, and having them validated by an outside firm.
They’ve collected a fairly impressive picture of the actual costs for supporting desktops for over 40,000 internal users across 61 countries. In addition to equipment costs, it factors in software applications, backup and archiving, the amount of server resources (and back-end storage, broken out by application), and even average lost user productivity due to outages caused by failures in desktops, servers, storage, networks, software, and so on.
The interesting part comes when trying to keep comparisons “apples to apples.” A simple “here’s today’s snapshot compared to tomorrow’s world” cost comparison is only valid if they provide the same value. But when the “new world” provides more value, it gets more complex. For example, if only a percentage of desktops are regularly backed up, comparing costs and benefits to a VDI infrastructure that provides backup and disaster recovery requires some thoughtful adjustments.
There’s also factoring in some currently unknown variables such as whether we consolidate the VDI back-end in a single continent or several, or what’s involved in regionalizing virtual desktops.
But even what we have now is a lot more detailed, and more reflective of a real-world scenario (well, ours at least) than anything we’ve seen publicly available. As soon as the current model is validated, I’ll share as much of it as I can.