The fact that I’ve been invited to a meeting to do with the design of the new building at work has set my mind on the task of how to persuade the head of department, the University’s Estates Directorate and the architects about the design for our new machine room. The problem is that this has been going around in my brain, so I’ve had to write it down,
Hence I’ve written an initial Keynote slidesshow detailing my thoughts, which are basically:
There is probably going to be a maximum of 4 clusters in the dept. at any one time, each being 2-3 racks in size. Each rack will weight about a tonne and output approximately 10Kw of heat, all of which will have to be taken away. So, that’s about 120Kw in total.
This is all OK so far.. now for the contraversial bit…
A conventional recirculation air conditioning system for 120Kw is BIG. Not only this but the failure modes when its near capacity are not graceful, with a catastrophic temperature rise due to thermal feedback. The systems would not be able to be shutdown quickly enough to avoid cooking of the disk drives in the room, leading to data loss.
My solution is based upon the premis that the external ambient temperature is less than or equal to the ideal machine room temperature for most of the year in the UK, so why not use that for cooling and vent the exhaust hot air out of the room? Not only this, if you design it correctly you can use convection to help the whole process (using a long enough chimney).
This idea has two advantages.. firstly you don’t have to cool 120Kw, merely tweek external temperature air to your required temperature, which should need far less than 120Kw. Also, if any one part fails the system fails gracefully. (If the chiller fails then the room maximum temperature will be slightly higher than external ambient. If the extraction blower fails then convection will keep at least some air flowing, allowing a slower heat build-up.)
I also suggested using a closed-loop ground heat exchange system for the air temperature control system, such as developed by a spin-off company from the Camborne School of Mines Geothermal Energy Project, GeoScience Ltd. and their subsidiary, EarthEnergy. This should be able to cut energy usage even further.
Of course, the exhaust warm air from the machine room could also be used as supplementry heating for the rest of the building as well.
Somehow, though, I think that the ever conservative Estates Directorate will veto any of these ideas, mostly because they couldn’t be implemented by their favourite contractors and the ever present “not invented here” syndrome.