And yes, it’s funny to see how they completely missed the big picture. I actually do believe that cloud computing will increase the computing budget of many companies, large and small. The reason is that suddenly compute resources become affordable and attractive to solve more business problems. Throwing some more machines or disks at the problem is cheaper than the alternative and suddenly becomes possible, even easy. So do it and save overall!

I read the physical server description – $14K price tag (2 CPU / 4 Core) – as describing a system with a total of 8 cores, not a 2×2 system. This math does pencil out to be ~$45/mo over three years (without depreciation), but it is still totally misleading. In my experience, enterprises frequently double (for on-site HA) or quadruple (for local HA, remote replication and remote HA) their CPU purchase for critical workloads. As you have noted, traditional I.T. shops also open their wallet for O/S licenses, HA capabilities (Veritas Clustering, VMWare HA, etc.), shared storage infrastructure, and, potentially, wide-area storage replication technology. Any comparisons that focuses exclusively on monthly price-per-flop is missing the boat.

In the wake of the McKinsey report, it’s certain that we will see a wave of more comprehensive financial comparisons; ones that accurately represent the costs of satisfying the functional AND non-functional (performance, availability) considerations in both traditional and cloud environments. This is what truly excites me.