Re:Why not both? (Score 1)

After reading many of the comments and replies I thought a little illumination on the power and cooling issues would be in order. First temperature and humidity. Theoretically, electronic components operate better at colder temperatures. In reality, both temperature and humidity must be kept in a very tight range for proper operation of servers. Too dry and you risk electrostatic discharge. Too humid and you risk condensation on circuit cards causing errors and failures. Second, cooling the room vs. cooling the servers. Regardless of the approach, the heat produced by the heat source must be removed. Most of the energy it takes to cool a rack of servers is used in rejecting this heat, not in cooling the surrounding air. Yes, energy is used by fans to circulate air across cooling coils, but other methods of heat rejection use similar amounts of energy to affect a thermal exchange. One watt of power consumed in a server results in 3.14 BTUs of heat energy that must be removed. It is the density of this power (and thermal) load that dictates the cooling tactics. High density racks loaded at 4kw per rack are typical and can be handled with standard air handling quipment. Very high density cabinets loaded at 10-15 kw per cabinet must use a combination of air and active thermal exchange at the cabinet level to achieve adequate heat removal. Third, AC vs DC to a rack. AC to DC power supplies are not a significant source of heat in the data center. CPU processors are the most significant and that is why multiple processor servers and blades require so much power. Spinning disk arrays in SANs and DASD cabinets are another. There is a growing inventory of switches, routers and servers available with DC power supplies driven by companies that come from the legacy telco world. However the DC suppies expect an input voltage of only 48VDC. A rack of servers requiring 4kw of power at 48VDC would need an electrical circuit of 100amps from a source less than 100 feet away to avoid voltage drop. These constraints result in a lot of wasted copper and higher critical infrastructure failure probabilities...not to mention doing very little to save energy. The reluctance of hardware mfgrs to go back to liquid cooled equipment will mean that we must deal with the growing power density issues with creativity and innovation.