Imagine, if you will, that you're planning a brand-new datacenter. Ah, the excitement! The anticipation! But oh, the costs ... Yes, they're going to be high, but at least you're now in an enviable position: You can determine your realistic anticipated needs for your new datacenter, which in turn will help you calculate how expensive it will be to construct -- as well as to power and operate after it's complete.
I emphasize "your realistic anticipated needs" because it might be tempting to overshoot and insist on more redundancy, cooling, and bells and whistles than you require. That is, until you realize just how significant an impact those extras will have on the datacenter's price tag. A recent white paper from The Uptime Institute, verbosely titled "Cost Model: Dollars per kW plus Dollars per Square Foot of Computer Floor," offers some perspective on those costs. As the title suggests, it provides a useful cost model to help organizations gauge how much they might expect as they budget for a new datacenter. ("Might" is an operative word here, as reiterated by the paper. Costs can vary significantly depending on projects.)
Your planning should begin by determining what tier datacenter you need. The Uptime Institute breaks down datacenters into four tiers: I through IV. A Tier I datacenter has a single path for power and cooling and no redundant components. It provides 99.671 per cent availability, meaning you can expect around 28.8 hours of downtime per year.
A Tier II datacenter also has a single path for power and cooling, but it does have redundant components, thus boosting availability to 99.741 per cent; around 22 hours of downtime per year.
At Tier III, you've got multiple paths for power and cooling to your IT equipment, though only one distribution path is generally active at one time. It, too, has redundant components. Availability for Tier III is 99.982 per cent, which translates to around 1.6 hours of downtime per year.
At the top of the heap is the Tier IV datacenter. This setup has multiple active paths for power and cooling, along with the redundant components and fault tolerance. Uptime is 99.995, meaning 0.4 hours of downtime per year.
The differences in projected downtime among these four models -- especially the jump between Tier II and III -- is indeed significant, and it may very well be tempting to decide that you must go with the higher number to spare yourself 20-plus hours of downtime per annum. But that's when it's time to take a hard look at the associated costs.
According to the cost model in The Uptime Institute's white paper, a Tier I datacenter could require an investment of US$11,500 per kW for redundant UPS capacity for IT. Tier II could require $12,500 per kW. Tier III, $23,000 per kW and Tier IV could be $25,000 per kW. (Again, I say "could" because of all the variables. The paper indicates the differences could be plus or minus as much as 30 per cent.)
On top of those costs, Uptime suggests that you plan to spend another $300 per square foot of computer room floor. You could also factor in $190 per square foot of empty space; that is, space set aside for a UPS system when your current system reaches maximum use.
The paper provides an example to illustrate how much of an impact moving from Tier II to Tier IV can make. A 20,000 square foot Tier II facility with 666 racks at 1.5 kW per rack (1,000 kW of UPS for your hardware) would cost a projected US$18.5 million. If you sought to make this 20,000 square foot space a Tier IV fault-tolerant facility, you'd have 666 racks at 3kW per rack (2,000 kW of UPS for your machines), which would cost a projected US$56 million.