It seems to be pretty much impossible to browse data center websites or magazines without being inundated with the challenges of data center cooling and the myriad of solutions being touted. If a data center professional made a serious effort to be well-read on all the topics, I think the data center itself would suffer from lack of attention. Personally I prefer the most simple and effective approach to just about anything I do. If I’m able to employ these two criteria, then not only is my day more productive, but the end result will meet my satisfaction.
CPI’s approach to data center cooling is simple and effective. I know, I’m adding to the noise by saying so, but the claim is worth investigating if you want a solution that is capable of taking your existing cooling capabilities and making it go farther – so far that you will use less cooling and power to get the same (or better) results.
The way we do it is simple. Our line of ducted equipment cabinets completely separates hot exhaust air from the cool air in the room. CPI pioneered the “chimney” cabinet which uses a duct to funnel hot air into the drop ceiling where it can be returned to the cooling unit. No more hot spots. No more recirculation. This concept, called CPI Passive Cooling® Solutions, has been tested and proven. If you have not had the opportunity to see how Intel incorporated CPI’s ducted cabinets, watch the video.
Let’s take it one step farther and talk for a minute about KyotoCooling® – a cooling system that uses a rotary wheel type of condenser to cool data center air without using a compressor. A recent article written by Rakesh Dogra on the Data Center Journal Website concluded that KyotoCooling seems to be the best possible answer to data center cooling thus far. What does this have to do with CPI Passive Cooling? Both solutions work hand-in-hand. Dogra explains the Kyoto concept like this, “The solution works in such a way that the cold air which is used to cool the hot air inside the data center does not actually mix with it … there are two circuits of air which are physically separated …”.
Even though CPI Passive Cooling does not rely on rotary wheel cooling, the end result is so effective that CPI is a regular presenter at KyotoCooling Seminars. The next series of seminars will be held December 1-2 and 3-4 in the Netherlands. Ian Seaton, CPI’s Technology Marketing Manager, will speak on “Maximum Cooling Efficiency with Hot and Cold Air Isolation at KyotoCooling: the Cooling Problem Solved”.
If the Netherlands is too much of a commitment, or if you have a question, contact us for more information by leaving a comment below or emailing techsupport@chatsworth.com. Kim Ream, eMarketing Designer
1. Humidity concerns always top discussion on free air cooling. First, since this blog chain has been mostly about KyotoCooling energy recovery wheel economization, I would point out that it is not a particularly serious concern since the actual exchange with outside air is less than 1/3 of 1% by volume as the heat exchange takes place through the revolving heat exchanger with separation between the data center and the outside world. In addition, in that environment, relative humidity can be controlled to a large extent merely by temperature modulation. As for traditional air-side economization, some form of relative humidity management is required. Fortunately, concerns about dry air and ESD are probably not founded. At this week's ASHRAE TC9.9 national meeting in Chicago, world-renowned ESD expert David Swenson delivered a very compelling paper on how low humidity is not a threat to our data centers. The basic premise was that if the equipment was properly grounded and otherwise prtoected, dry air would not cause an ESD threat and, conversely, if the equipmetn was not properly protected, moist air would not protect the equipent from ESD. While the resultant investigation proceeds now, there is still a low dew point spec in the ASHRAE 2008 Environmental Limits, but it is being defanged and high humidity is going to be where the concern resides and some form of ultrasonic control is typically going to be the most efficient.
2. Most servers will see their internal fans start to kick up and consume more power when the inlet temperature exceeds 77 degrees (25C), which means there is now a 3.6 degree window of extra fan energy consumption inside the new ASHRAE environmental limits. So first, you can drive your inlet temperature past that 21C (70F)without incurring any penalty from your server fans. Above 25C, you can start paying that penalty, and we all know how that is decidedly not a linear relationship between airlfow and energy and we will be the loser. However, you need to weigh that against the benefits of the higher temperatures, particularly if you are running free cooling. The proportional fan energy budget for a 1U server is much lower than it is for a typical blade center, so the first thing is to determine exactly what that is compared to the energy budget for the chiller plant, which can be 75% of the total cooling budget or 35% of the total data center energy budget, and then determine which is costing you more. I would say that you would normally expect the savings for not running the chiller plant to exceed the extra cost of running those server fans faster; however, if the only benefit you're looking to get from running your inlet temperatures higher is a chiller plant efficiency benefit of around 1% per degree, then the extra 3 degrees at the higher server fan levels is not likely to have a positive payback.
Posted by: Ian Seaton at 1/30/2009 3:12 PM
|