What is a data center was defined as follows:
• Any room that is greater than 500 square feet and devoted to data processing; and,
• Meets one of the tier (I, II, III & IV) classifications defined by the Uptime Institute.
The problem with the room size is that with modern server technologies the space required for housing large computing power is only minimal. For instance a single rack-mounted IBM eX5, with mainframe computing power, will occupy only a fraction of space in a 20x25 room.
Defining a data center as meeting Tier I classification criteria would qualify installations that are currently not included in Kundra's count. Tier I installations do not require a raised floor and do not need a source of uninterrupted power supply. Tier I data centers operate with only a single system and have no redundancy. They have multiple single points of failure. A Tier I data center can be placed almost anywhere in a temperature controlled office environment or in a shipping container.
In the last two years Sun Microsystems, Hewlett-Packard, Dell, Microsoft and SGI have offered complete data centers in standard 40x8 ft and 20x8 ft shipping containers, occupying 320 and 160 square feet respectively. Such data centers have the capacity of up to 29.5 petabytes of storage and up to 46,080 CPU cores of processing power. These data centers would not be included in any of the OMB surveys!
Perhaps the greatest omission in the DoD data center count are installations that are operated by contractors. For instance, the data centers for NMCI, which are owned and operated by HP/EDS, are not included. That could understate the amount of computing dedicated to the support of DoD.
The DoD count of what is defined as a “data center” most likely understate what is the actual number of operations that should be considered for consolidation. Current rack space technologies offer a large multiple of the computing power than was available in much larger configurations less than ten years ago. With oversight dictated by project budgets, the urgency of short implementation schedules will dictate the installation of powerful computer configurations unaccounted for by the Federal Datacenter Consolidation Initiative Report.
From the standpoint of cost, economies of scale, pooling of resources, more effective use of personnel and the concentration of security expertise the DoD data center count needs to be revised before it can serve as a basis for long range planning.