So finally someone comes up with a cooling solution that makes perfect sense to me…and implements it.
Everyone is talking about the SuperNAP by Switch Communications that is supposed to be one of the largest and most efficient datacenters produced to-date, beating the likes of Microsoft and Google in terms of capacity per square foot.
Everyone is linking to Ashlee Vance’s articles in The Register (here and here) about 8-year old Switch Communications which has gone unheard of for all these years until now. The gist is that they aquired Enron’s pipe-dream of “bandwidth as a commodity” datacenter for next to nothing and in doing so inherited the ideal foundation for colocation services. They picked up military contracts and other, shall we say “big-time, low-key” customers, and built a business on it.
I have to say, I like what I’ve seen so far.
The two things that stand out are the use of ex-military personnel for security, the very presence of the actual military in the building, and their method of cooling, part of their success in the capacity-per-square-foot department.
In most datacenters the onsite security amounts to nothing more than a (sometimes) convenience to customers, scarecrow for security, and an eye witness account of what happened but couldn’t be prevented. Ex-military with big guns tends to produce a different outcome. Burst through a wall with your little tazor and be ready to be blow to Kingdom Come by the datacenter Rambo. The presense of the actual military gives you the upper in hand in early warning for world-wide events and threats.
And finally, with respect to cooling scenarios that never made any sense to me, we finally have a company designing a front room and a back room — it was about time. The “front room” provides access to the front of the cabinet where you insert your servers and (I assume) plug any gaps with filler plates. Cold air is pumped into the room from above and is completely isolated from the rear of the cabinet. The “back room” is dedicated to managing hot air that is collected from above and cycled through the system. Brilliant, if not obvious.
To top it off, they have four (4) different methods of cooling available to them, and they choose amongst the four (sometimes using two at a time) to most efficiently cool the equipment. Again, brilliant, if not obvious.
The last bit to catch my eye was the use of color. They color code the equipment and locations and then designate which color can be worked on during a given period. All of the equipment is redundant and each component of the redundant system is in each of the colors. Therefore they reduce the chance of human error by controlling which color can be maintained on that day. Brilliant, again.
In any event, it makes a really good read and I would highly recommend Ashlee’s articles. I can’t think of a better place for mission-critical remote desktop solutions. Let’s see who else can catch up to Switch now that the bar has finally been raised.