Rich - Nice analogy! (Refrigerator vs walk-in cooler) -TS
-----Original Message----- From: tech-geeks-boun...@tech-geeks.org [mailto:tech-geeks-boun...@tech-geeks.org] On Behalf Of Richard Mann Sent: Saturday, October 02, 2010 5:30 PM To: Tech-Geeks Mailing List Subject: Re: [tech-geeks] Server Heat Load I think the server load calculator I've been using is at http://www.apc.com/tools/ups_selector/index.cfm. I can echo some of the experience in this thread. If I could do it all over again {begin dream ripple ~~~} I would have all closed-type cabinets with closed-loop A/C (as TS alluded to). It's like at home. You don't put your cold food on open shelves in the kitchen, and chill the whole room, do you? But most data centers use that model. A closed rack/cabinet with dedicated A/C (i.e. a "refrigerator", in the kitchen analogy) is going to get the cooling where you need it, do the best job, and save a bunch of money compared to leaving things in the open. Day after day after day. In a data center, some achieve this with a raised plenum floor, and a ceiling plenum return. The cabinets connect top and bottom. One thing we did do right at ICN, if you are also talking high-reliability, was split power across two systems. (This was a tip from UIUC.) Everything's got dual power supplies, so we have an A/B power system and one each PS plugs into A, the other B. HVAC is also dual units. I think myself I'd usually still prefer 120VAC on the output side, so I can easily swap in off-the-shelf hardware. But we do have both 120 and 208 readily available for ICN equipment. Better end the dream ripple, here we go, {~~~}. It's still the weekend at least. Rich / ICN-RTC7 Champaign On 10/2/10, Steele, Thomas C <tste...@manteno5.org> wrote: > With UPS you have to consider both the power consumption in standby mode and > in operational mode. Typically, though, unless you have a generator, you > won't have any HVAC running when the UPS is in operational mode (and if you > have a generator, your equipment shouldn't be running off the UPS) so that > probably won't be a factor for most schools. Related to UPS you also need > to factor in any PDU's - again for most school settings this would be > negligible, but they do generate a slight amount of heat. If you have a > system that also provides line conditioning, that, too will generate extra > heat. I saw a formula - I think from APC - for calculating the power/heat > load from UPS/Power conditioning equipment. > > Higher voltage is slightly more efficient, but again, in a typical school > environment the difference would be negligible. A central DC feed is, > indeed, more efficient in a data center because it eliminates the need to > convert power individually in each piece of equipment. Each power supply > consumes power in the process of conversion, thus generating more heat - > directly fed DC eliminates the need for these extra power supplies. I have > not seen too many servers of the type used by schools that have this option, > but the Cisco gear has had the option for DC power for quite a while. > > If truly trying to get as accurate as possible, you can also subtract any > power used by PoE devices from the cooling load since the heat is dissipated > by the device which is most likely outside of the data center. Based on > current PoE specs, this would typically be anywhere from 1W per device to > 15W per device. > > One other thing... if the electrical engineers decide to put step down > transformers inside the data center, you will have a whole lot more heat to > dissipate. We were fortunate enough to have that happen in two of our > network closets. > > If you want to go with the most efficient design, localized cooling inside > the rack with exterior venting is the best option (if you have the design > flexibility). > > -TS > > From: tech-geeks-boun...@tech-geeks.org > [mailto:tech-geeks-boun...@tech-geeks.org] On Behalf Of Jim Flanagan > Sent: Saturday, October 02, 2010 11:02 AM > To: Tech-Geeks Mailing List > Cc: Tech-Geeks Mailing List > Subject: Re: [tech-geeks] Server Heat Load > > Also, don't forget the UPS's. > > And, if doing a big server room, other's more knowledgable please correct > me, but aren't their advantages to be had in minimizing heat Load by looking > closely at how you power everything. > > I don't remember the details but 110AC I think is the least efficient. One > one end 220V and at the other DC, although I guess DC is bleeding edge. > > This site looks really useful for reference materials and design tools > > http://www.thegreengrid.org/library-and-tools.aspx > > Jim > > > > Sent from my iPad > Cell 773.396.4638 > Chicago 773.463.5494 > Beaver Island 231.448.2109 > > Privacy Notice: The information contained in this electronic message is > intended only for the use of the individual or entity to which it is > addressed. This message is private and may contain information that is > privileged, confidential and/or inside information and exempt from > disclosure under applicable law. If you are not the intended recipient, you > are notified that any dissemination, distribution or copying of this > communication is strictly prohibited. Please contact the sender by reply > e-mail and destroy all copies of the original message > > On Oct 2, 2010, at 9:41 AM, "Steele, Thomas C" > <tste...@manteno5.org<mailto:tste...@manteno5.org>> wrote: > We had to do this during our 2007 construction project. You could get > extremely technical, but for the most part a simple calculation will be > sufficient. > > > 1) Add up the total rated watts drawn by all IT equipment (servers, > switches, etc). If the wattage is not given, multiply the Amps x Volts > (i.e. if it is rated at 2A at 120V, multiply 2 x 120, resulting in a power > draw of 240W) > > 2) Multiply the total watts by 3.41 to arrive at the total BTUs/hr > generated by the tech equipment > > 3) To convert to tons, divide the total BTU's by 12000 > > 4) This represents the ADDITIONAL cooling required by the IT equipment > above and beyond what would otherwise be required for the room. > > You might be shocked by the results. It turns out I have more cooling in > our 8x10 server room than I have in my whole house! Also, when doing your > calculations, try to anticipate future needs. A rack full of 1RU servers > will generate a whole lot more heat than a rack of 5RU servers so consider > what you *might* eventually have in the room. On a related note, if you > have to size AC for a computer lab, follow the same formula, but add an > additional 100W per person to compensate for additional body heat - this is > something most architects/engineers seem to overlook when designing labs. > It cracks me up when I see people trying to "cool" their labs using 5000BTU > window air conditioners - do the math! > > Thomas C. Steele > Technology Director > Manteno CUSD #5 > 250 N Poplar > Manteno, IL 60950 > ph: (815)928-7006 fx: (815)468-6439 > > > > From: > tech-geeks-boun...@tech-geeks.org<mailto:tech-geeks-boun...@tech-geeks.org> > [mailto:tech-geeks-boun...@tech-geeks.org] On Behalf Of Egolf, Ron > Sent: Saturday, October 02, 2010 8:46 AM > To: tech-geeks@tech-geeks.org<mailto:tech-geeks@tech-geeks.org> > Cc: Richard Mann > Subject: Re: [tech-geeks] Server Heat Load > > ICN had to do this when we started racking up the LTC cloud. Give Ben or > Rich an email call. > > > On 10/1/10 11:16 AM, "Eric Roemer" > <roem...@woodland5.org<mailto:roem...@woodland5.org>> wrote: > Anyone ever had to calculate the heat load for your server room? We are in > need of upgrading the cooling system in our server room and our AC vendor > has asked me to get him the heat load from my servers. I have 10 servers in > the server room. Any info on getting a quick, usable estimate of the heat > load of the servers would be very helpful. > > Thanks! > Eric > > THIS TRANSMISSION IS INTENDED AND RESTRICTED FOR USE BY THE ABOVE ADDRESSEE > ONLY. IT MAY CONTAIN CONFIDENTIAL AND/OR PRIVILEGED INFORMATION EXEMPT FROM > DISCLOSURE UNDER FEDERAL OR STATE LAW. IN THE EVENT SOME OTHER PERSON OR > ENTITY RECEIVES THIS TRANSMISSION, SAID RECIPIENT IS HEREBY NOTIFIED THAT > ANY DISSEMINATION, DISTRIBUTION, OR DUPLICATION OF THIS TRANSMISSION OR ITS > CONTENTS IS PROHIBITED. IF YOU SHOULD RECEIVE THIS TRANSMISSION IN ERROR, > PLEASE CALL WOODLAND CUSD#5 IMMEDIATELY AT 815-672-5974, DELETE THE FILE > FROM YOUR SYSTEM, AND DESTROY ANY HARD COPIES OF THIS TRANSMISSION. THANK > YOU. > > THIS TRANSMISSION IS INTENDED AND RESTRICTED FOR USE BY THE ABOVE ADDRESSEE > ONLY. IT MAY CONTAIN CONFIDENTIAL AND/OR PRIVILEGED INFORMATION EXEMPT FROM > DISCLOSURE UNDER FEDERAL OR STATE LAW. > IN THE EVENT SOME OTHER PERSON OR ENTITY RECEIVES THIS TRANSMISSION, SAID > RECIPIENT IS HEREBY NOTIFIED THAT ANY DISSEMINATION, DISTRIBUTION, OR > DUPLICATION OF THIS TRANSMISSION OR ITS CONTENTS IS PROHIBITED. > IF YOU SHOULD RECEIVE THIS TRANSMISSION IN ERROR, PLEASE CALL WOODLAND > CUSD#5 IMMEDIATELY AT 815-672-5974, DELETE THE FILE FROM YOUR SYSTEM, AND > DESTROY ANY HARD COPIES OF THIS TRANSMISSION. > THANK YOU. > | Subscription info at http://www.tech-geeks.org | > | Subscription info at http://www.tech-geeks.org | > -- * In a world without walls, who needs Windows, really?* Charles-H. Schulz of The Document Foundation | Subscription info at http://www.tech-geeks.org | | Subscription info at http://www.tech-geeks.org |
