Apologies for a late summary. Thanks to everyone who replied. There was no concensus on what to do. I am going to post a few comments below: Ian: " Easiest thing to do is get the electrician to come in and take a current reading on your current room and then add some padding based on the difference between the room sizes." comment: we did just that. an electrician came in, told me what we have ("Tree Phased 60Amp panel w/ isolated ground and 18x20Amp circuit breakers inside"). The actual usage was 27Amps (he took readings), and I've told the architects/engineers that we'd want a similar setup at the new place, we plan to decrease the amount of hardware anyway (place is tight) Chris: "I don't think those numbers are unreasonable. If you try and chop those figures down you are taking on a liability that you probably don't need. Are you concerned about the money? If it was me I would ask for a 200 amp service and AC to provide 100K BTU/hr." comment: yes we are concerned about the money. this is also a development environment, and we are not too concerned if it goes down. Gary: "Are those numbers calculated from tags on back of hardware or Suns System Handbook? My experience is that the power usage/heat numbers from the handbook usually are smaller than the max numbers calculated from the hardware tags. Look at: http://sunsolve.sun.com/handbook_pub/ " comment: We noticed the difference, but it seemed insignificant. Anjan: "3 half full racks at 110Amps is the MAX rating you are looking at. APC has a good configurator you can use." commment: I played around with it, but it seemed to cumbersome. Plus I figure they'd be biased to peak usage, to upsell higher end models. Thomas: "What do you do for UPSes, or is that designed into the server room? We use APC stuff generally, and they have some good specs for power ratings of various common devices. I don't know how unrealistic your numbers are. You should also check what the construction people want --- do they want maximums, or averages (in which case presumably they tack on some extra for maximums). I would think your current ratings are not unreasonable; max load tends to occur at start up for a lot of devices (fans, disks draw more power as they get up to speed), and I assume you want enough available power to start everything all at once (even if never planning to do such). Our server room is not huge (3 x Sun280R servers, a Cisco4006, about 6 alpha workstations, a few disk shelves, and some x86 servers, and we have one dedicated 100A line to one UPS, another big UPS with a dedicated 30A line, and about 4-5 20A circuits in the server room. Admittedly, not that close to maxing out, but I would ask for a lot of power. Also remember, if dealing with UPSes, they use VA as the unit, and are concerned with peak VA, which is about 40% greater than wattage of the device (because of the sinusoidal wave form of the AC current). " comment: construction people are a bit difficult to deal with. I don't think they were ever involved in building a server room. We actually would be OK if we didn't start everything at once. Richard: "If you find such info please share - working on a new server room and the "listed" vs "real" issue is several hundred thousand dollars. by actual measerment with ampmeters, I get a real load of 37% to 62% depending on cpu/ram useage. if you use 65% overall you should not trip breakers, and not break the bank. I would not go below 40% for a working system, and would expect to trip breakers occasionally at that level." Daniel: "You might be interested in an exercise I have made. These are the consumptions that I have found out (240V supply)Your consumptions might vary according to your environment (ie usage and bla bla bla ) so don't just paste them in your report Model Maximum stated current Power(W) (Supply type Dual / Single) Current Consumption (All supplies together) Netra T1 model 105 2 108 Single 0.6 E220R 6.3 610 Dual 1.3 D1000 2.6 260 Dual 1 E450 13.8 1664 Single 2.81 " Mark: "I don't have references to send you, but the empirical numbers we use are 5 kVA and 15000 BTU/hour for each fully loaded rack. This is assuming a good mix of 1U / 2U / 4U servers, plus several fully populated A1000 or similar RAID arrays. We have found these numbers to be pretty accurate in our environment. Note that if you're doing blade computing or clustering, this can significantly increase both of the numbers above. " comment: I've heard similar "rule of thumb" rules. Best is probably to call a data center and ask what they do. Wendell: "I give our facilities guys the max current draw for each server so they can balance out the load across the power source. If the power fails, its possible that a value close to max power for all your servers draw might be needed for a brief moment of time when things snap back up, make sure they give you enough in-rush power capacity. If you give them your in-rush needs on all system startup you will be fine for normal operation. Don't worry about the BTU/hr in relation to max current draw, they want BTU/hr for cooling. Just let the HVAC techs know your max BTU/hr for all your servers that way they give you enough cooling. Don't forget to fudge the numbers so you can add a rack later! " comment: I let the engineers worry about AC, plus I think they'll just put in a similar unit to what we have now. As for 'growth', we actually plan to shrink. Having 15 servers supporting 15-20 people is a bit too much, in my opinion On Thu, 22 Apr 2004, NetComrade wrote: > Gurus, > > We are moving offices, and the construction company is bugging us about > power and AC requirements for a new server room (we have 3 half full > racks) > > We have looked at the servers, and e.g. an E450 has a maximum power usage > of 13.6Amps (I would guess at 100V), and maximum heat output 5680 BTU/hr. > > However, how do I get these #'s down to a realistic #? > > Is there are some rule of thumb? If I calculate all maximum powers of all > equipment I get something like 110Amps and 47000BTU/hr, which is highly > unrealistic. > > If you can point to some docs on standards, that could help too. > > Thanks. > _______________________________________________ > sunmanagers mailing list > sunmanagers@sunmanagers.org > http://www.sunmanagers.org/mailman/listinfo/sunmanagers _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Tue May 18 11:59:33 2004
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:32 EST