What do you think about self-cooling racks?
I surmise that you are referring to what have commonly been called "Water Cooled Racks" – cabinets that have water jackets and/or refrigeration systems built into them in one way or another. This would be as opposed to some of the more sophisticated "air flow enhancers" for cabinets such as the Adelphi or APC Blowers, Wright-line's "Tower of Cool," or the various rear-door fan systems available from a variety of manufacturers.
It would be inappropriate for me to give you a specific opinion on any manufacturer's product without being able to publish controlled test data to back it up. What I can do is provide a set of considerations and question (not necessarily complete) you can use before deciding on one of these specialized cabinets, or on the approach as a whole.
The most important thing to remember when considering the different ways to cool data hardware, is how the necessary amount of air, at the appropriate temperature, will be moved through the equipment. Every quality piece of modern computer technology has been designed and laboratory tested to ensure that the critical components (primarily the processor chips and Memory, but also things like Disc Drives) can be adequately cooled by the internal fans IF enough air is made available to the unit at the right temperature. Manufacturers who follow the ASHRAE recommendations for reporting power and cooling requirements (unfortunately far too few, but growing) provide very thorough and accurate data in this regard. (Note that the ASHRAE format also states the maximum air flow and lowest temperature – something to keep in mind with all types of refrigerated cabinets and air boosters.)
The point being made here is that simply putting a piece of computer hardware in a refrigerator will not solve the cooling problem. Lowering the case temperature by surrounding the device with cold air may help remove internal heat to some extent, but it will not, in and of itself, properly cool the internal components. To cool the innards requires getting that cold air to the intake ports (the front of the device according to industry standards) so the internal fans can pull it through in the way the machine was designed. Therefore, the first question to ask the manufacturer of any of these "closed cabinet" systems is, "How do you make the cold air available to the device?" Notice that I did not say, "How do you pull the cold air through the device." Huge fans pulling cold air through the equipment can create a number of problems inside. If the amount of air drawn through the case greatly exceeds that which the internal fans were designed to move, then the internal fans become useless and air flow inside the computer case may not be what was intended, with air actually being diverted from the places that need it most.
The next critical question is "How to you maintain cooling below the dew point?" The dew point is that temperature at which the moisture content, or humidity, in the air condenses back into liquid rather than vapor form. Water condensing on internal computer components and dripping through them is obviously bad, and there are only two ways to avoid it; lower the humidity or increase the air temperature. Lowered humidity invites static, which can be caused by high air velocities as well as by other means. Increased air temperature means requires more air volume through the computers to achieve the same amount of heat removal. Do not be surprised if, with some thorough investigation and computations, you discover that the advertised heat dissipation of the cabinet can't be met without resorting to a lower temperature level or higher air velocity than the computer manufacturer recommends, or a lower humidity level than is deemed prudent. Most of these cabinets are advertised as "Up to XX,000 Watts" for a very good reason.
In investigating these cabinets, you should also look very closely at the redundancy and ease of replacement of the various cabinet and cooling components, asking yourself and the vendor in every situation what would happen if this failed, or that failed, or if the cabinet door is opened, or if power fails, etc, etc, etc. (As just one example, consider what happens if both the internal cabinet temperature and humidity must be kept very low to achieve the advertised cooling capacity, and a cabinet door is opened. Instantly, components are exposed to the significantly higher humidity level maintained that is probably maintained in the rest of the Data Center environment. Moisture may quickly condense on and in the computing hardware, just as it does when you've been in an over-cooled space like a grocery store in summer and then step into the humid air outside and see your glasses instantly fog.)
Lastly, evaluate the accessibility inside the cabinet, for things like installing and removing hardware, installing plug strips, dressing cables (permanent cables, patch cables and power) without blocking equipment air exhausts, and seeing inside to read markings and to access connectors. Compare this accessibility with that of standard cabinets (assuming you're using cabinets of sufficient size already) to see whether the "self-cooled" cabinet you are considering will hinder your daily work.
The one personal opinion I can give regarding these cabinets is that I try in every way possible to design so that a facility doesn't and hopefully won't require these kinds of special solutions. I also believe that we will see more solutions in the very near future that bring cooling of one kind or another directly to the server. With the way power levels and, consequently heat loads and thermal densities, are increasing, this will ultimately be a much more efficient and practical way to cool high performance devices than trying to direct air into and through them. (Notice that the one thing I did not voice concern about was the need for water lines in the data center to support any of these devices. Consider it inevitable.) Until then, we may need to employ self-cooling cabinets, as well as any of the other techniques currently available to us, in order to keep this very expensive hardware in operation. But when we utilize any special approach, we must do so knowledgably. If that expensive high-end server fails, its manufacturer won't really care what the cabinet salesman said his or her approach would do.
