Most data centers are inefficient, in the wrong place and built to the wrong scale. It’s sad, but it’s true and IT departments across the world are going to have to adjust to this reality. It’s an unwelcome surprise and right now there’s a certain amount of “cloud denial” taking place.

Nevertheless, the destiny of the data center is clear. It’s going to evaporate into the cloud and it will happen a lot faster than many people expect. What I’m talking about is Infrastructure as a Service (IaaS) which looks set to outpace other cloud propositions such as platform as a service and software as a service (PaaS and SaaS).

Why?

Let’s cut to the chase:

For a well scaled data centers with tens of thousands of servers, you can roughly categorize annual data center costs in the following ratios as:

• Land 2%
• Architectural (i.e. buildings) 7%
• Core and shell costs 9%
• Mechanical/Electrical 82%

(for source click here)

The main point here is that 82% of the costs are directly proportional to the computer workload and less than 10% are proportional to building space. And this means that if you’re hiring out data center resource you may as well charge by the watt of electricity consumed.

There’s an interesting Microsoft paper that you can download, which discusses the issue in some depths – using a hypothetical 50,000 server data center. It gives the following electromechanical cost ratios.

• Server hardware: about 45% of the cost.
• Infrastructure: (i.e. power distribution and cooling.) 25% of the cost.
• Power (electricity cost): about 15% of the cost.
• Networking equipment & comms: 15% of the cost.

Can you see what’s missing here? No?  How about staff? Microsoft didn’t include staff costs, because it was less than 5% of the total. They suggest that a scaled up data center needs about one member of staff per thousand servers, compared to one per hundred in the typical corporate data center. If you merge the two sources I’m using here, you get these percentages:

• Land 2%
• Architectural (i.e. buildings) 7%
• Core and shell costs 9%
• Server hardware cost 36.9%
• Power distribution and cooling 20.5%
• Electricity cost 12.3%
• Networking equipment & comms: 12.3% of the cost.

The Bottom Line

So what’s the bottom line? The actual cost ratios between these various items don’t really matter. I’ve seen figures that suggest higher power costs than these and lower server costs. These ratios will certainly move around a bit depending on what kind of workloads you happen to be scaling, local power costs and other contextual factors. But what matters most is that the staffing costs are almost irrelevant.

Question: What kinds of businesses are there where staffing costs are pretty much irrelevant?

Answer: Oil & Gas, Mining, Steel, Chemicals, Utilities, Heavy Construction, Shipbuilding, Heavy Engineering, etc.

What makes winners and losers in these industries is effective utilization of very expensive assets. That’s what sorts out the heroes from the zeroes.

Data centers as they exist now “in the enterprise” are a cottage industry that is about to be gobbled up by mass production. If you want to know what’s going to happen, then think in terms of the “cost per watt” of a data center falling by about 80 to 90% in the next 10 years due to massive scaling. Historically, that kind of cost collapse can be pretty much guaranteed when a cottage industry gets replaced by truly efficient mass production.

That kind of cost collapse guarantees that data centers are going to evaporate into the cloud.

See also:

The Death of the Data Center: The Model
The Death of the Data Center: Location, Location, Location
The Death of the Data Center: Power
The Death of the Data Center: Cooling
The Death of the Data Center: Networking
The Death of the Data Center: Server Hardware
The Death of the Data Center: The Software
The Death of the Data Center: Software Optimization