Doing its best to imitate The Fonz, Dell has formed an exclusive partnership with data center cooling specialist Liebert to put your boxes on ice. Okay, sure, The Fonz wouldn't have wasted his time talking about CRAC units and maximizing air flow. Server vendors, however, sure think cooling data centers is hip. So, we find Dell …
A bit puzzled by the claims. The heat has to be removed somehow, and heat transfer requires an energy input. How does this system of refrigeration work so much more efficiently than a regular air conditioning system? At the end of the day, a bunch of warm air is cooled and the waste heat is rejected. Why would a bunch of little-bitty air conditioners be more efficient than one big one?
I understand that they're using a single-point compressor and distributed heat exchangers, but I still don't see where the savings are. Dumping the waste heat into an external sink (groundwater, a convenient river nearby, evaporative cooling tower, etc.) would improve efficiency, but that doesn't appear to be what they're doing.
Say , who failed basic laws of Thermodynamics , that be the question , or has Dell finally created a new universe where the laws of physics of the real world do not matter!
Re: Energy savings?
I think that the reason you can make it more efficient is that it's easier to cool air at 40 degrees down to 20 than it is to cool air at 20 down to zero, yet both remove the same amount of energy.
The reason for this is that if your refrigerant after evaporation is at -10 (or whatever), the temperature differential is much higher if you're passing the hot air straight from the server over it, so the heat is absorbed more quickly, and also more of it.
By analogy, it's more efficient to boil a saucepan of water if your cooker ring is at 200C than at 101C.
Interestingly most A/C units have their capacity quoted at 30C / 80% RH, which means that in the UK at least, that's rather optimistic most of the year!
It *can* work... but only if done right.
I would need to see more details to say if it can, but its quite possible to get a reduction in cooling power consumption this way. If it reduces the temperature difference between ambient air and the components (Which watercooling will do), it allows the datacenter to be kept a couple of degrees warmer, which reduces demand on the aircon. In addition, by directly coupling the coolent water to the aircon system via a heat exchanger, the intermediate coolent of amibent air can be done away with - again, this results in a reduction in the temperature difference, and thus even more savings.
That said, its really easy to screw this up. You cant just stick basic PC-cooling parts together with some hose and a pump and expect it to work right. It needs an expert to work out flow rates, resistances, how to ensure it works even if one pump fails (You certinly want redundency!), specialist self-sealing hot-plug plumbing fittings so you can take servers out without spraying water everywhere...
Why can't you just use fresh air?
.. lots of it. In the UK, the ambient temperature rarely rises above 30C, so you just need big ducts with big fans. No heap pump needed.
Did you look at the Liebert literature?
Having a mechanical engineering background, I could see the efficiency increase immediately upon looking at the literature. The system is designed to connect to a buildings main chiller system (IE Cooling Tower) as well as stand alone. Since the main chiller system of a building is much more efficient than the "stand alone" Computer Room AC units, the new system offered by Liebert can take advantage of existing building cooling capacity, closely couples the AC heat exchange units to the server racks (less loss), and directs airflow where it is needed, not all over the room. Other systems are the equivalent of sticking an AC unit in the wall 30 ft from the heat load.
By the way I don't work for Liebert.
The only process that might be more efficient would be to have refrigerant based cooling that is directly coupled to the CPU heat exchanger. But that would be difficult to work with.
For what it is worth Suricou, it is coolant, not coolent.
Efficiency claims will eventually be validated
I have a lab of these cooling units at Dell. As soon as I'm able to measure the chilled water used by the entire XD system, I'll be able to verify the efficiency claims that do originate from Liebert. In theory, here is where the efficiency advantages come from: by being located closer to the racks (either hanging in a cold aisle or directly on top of the rack), the fan energy required to draw air across the cooling coil is much less than it is to push a comparable amount of air through a raised floor air handler and through the raised floor to the vent in front of the rack. The coil units overhead are provided liquid refrigerant via remote heat exchange / pumping units. The system is tuned such that the heat the coil picks up from the IT equipment creates a phase change of the refrigerant to gas as it absorbs the heat. This method of heat transfer is extremely efficient and requires less refrigerant than would a water based heat exchanger that operates as a single phase heat exchanger remaining a liquid. The pumps in the XD pumping unit are smaller and require less energy because they are pumping less fluid. The other efficiency advantage is that the XD system is designed to operate above the room's dew point all the time. There is no condensation in any of the XD coils. Condensation and subsequent re-humidification, which is common in CRACs and CRAHs, is a big energy drain.
And yes, if you happen to be fortunate to be located in a climate that enables fresh air cooling, that would generally be even more efficient. ASHRAE TC9.9, the group responsible for environmental recommendations is looking at the possibility of enlarging the recommended humidity ranges for IT equipment. If this happens, more of us can consider free cooling.
Link to Whitepaper
Here's a link to the paper that David Moss authored about this solution ... http://www.delltechcenter.com/page/DataCenter+Power+and+Cooling
- Geek's Guide to Britain INSIDE GCHQ: Welcome to Cheltenham's cottage industry
- 'Catastrophic failure' of 3D-printed gun in Oz Police test
- Game Theory Is the next-gen console war already One?
- Analysis Spam and the Byzantine Empire: How Bitcoin tech REALLY works
- VIDEO Herschel Space Observatory spots galaxies merging