How to Remodel Your Data Center

The Strominator gets off on data center architecture, implementation and cost savings.  

I have to come clean: I am a data center geek.

I love visiting them, and talking to the people who design and run them. Maybe it was because I worked around mainframes at the beginning of my IT career. Maybe it is all that power coursing through all those wires, and those big, greasy generators. Or the thrill of getting access to the inner sanctum of IT after passing through various security checkpoints and 'man traps.'

There is just something about a hyper-cooled raised floor that gets me excited. Okay, enough of that. But there are some interesting things happening in data centers, including companies that are trying to downsize them to save power and money.

Take a look at what Sun did last year, when they brought up a new datacenter in Santa Clara that was about a third the size of the old one. Through various cash awards and utility rebates, they saved a million bucks right off the bat, and that doesn't take into account the monthly operating cost savings, which they estimate were about 60 percent of their old electric bills and will bring a three-year payback return on their investment.

And lest you think this is just a one-shot deal, Sun's goal is to have no datacenters by the year 2015, at least according to this blog by Brian Cinque, Sun's data center architect.

Now, an unkind response to this notion might be the reason they will have zero data centers is because the company might be out of business by then, but let's give Sun the benefit of the doubt and credit for their achievements, nonetheless.

So, how did they do this? Several ways: server consolidation, server virtualization, replacing older units, and increasing rack densities.

They started by grouping a bunch of racks, servers, and other equipment together in what they called pods that would be designed with the same requirements and could easily be scaled up in the future.

Didn't I say they were downsizing? Yes, but they are doing so by cramming more gear into smaller spaces, and replacing older space-hogging hardware with newer gear that is more compact. The racks in each pod were also designed to handle at least 5 kW now, and be scalable to 18 to 30 kW in the future to allow for even higher density deployments per rack.

The pods allow Sun to deploy modular solutions for cooling, cabling, monitoring, and power, and to do so independent of the vendors needed for the gear in each pod. They took advantage of APC'x Infrastructure line and Liebert XD equipment, and other vendors as well. You can see more explanations by Sun's own staffers in a nice series of videos.

Sun ended up tossing about 5,000 servers and switches as a result of the move. Well, they didn't actually toss all of this – they tried to recycle as much as they could. But their approach is worth taking a closer look, especially if you are in the market for a new data center, or want to start exploring ways that you can cut your own costs and make your operations more efficient.

Next week, I will talk some more about what APC is doing to help data center consolidation and a visit to their data center test bed.