For planning purposes I just did some power draw testing of a Dell PowerEdge R610. Dual Intel X5550 CPUs, 24 GB of RAM, four SSD disks attached to the PERC6/i, and dual 717 Watt power supplies. My testing methodology was to measure the draw using a Fluke 322 clamp meter, both at idle and running a stress test under Red Hat Enterprise Linux 5 (stress -c 32 -d 8 -i 8 -m 16). I did this with one and two power supplies active.
1 PS, idle: 0.65 Amps @ 202.3 Volts = 131.5 Watts
1 PS, loaded: 1.51 Amps @ 202.3 Volts = 305.5 Watts
2 PS, idle: 0.35 Amps @ 202.3 Volts = 70.8 Watts each (total of 141.6 Watts)
2 PS, loaded: 0.77 Amps @ 202.3 Volts = 155.8 Watts each (total of 311.6 Watts)
Virtualization users who didn’t see the VCritical commentary on “Idle RHEL Hypervisors save power?” might want to check that out, since these numbers directly support Mr. Gray’s argument. An idle server drawing 45% of the power of a loaded server is a pretty solid argument for VMware DPM.
Also note that the label on the power supplies indicates these are 717 watt power supplies, while the maximum draw I recorded was less than half that. While building in some overhead is a good idea, using the labels to determine draw isn’t a great idea, because you will overbuild your infrastructure.
Update: As was pointed out in the comments, you can use the Dell iDRAC web interface to find out the current power consumption of the 11th generation Dells. However, it looks like that has accuracy problems, particularly at the low end (idle), but is relatively usable at the top end. If you’re serious about measuring this stuff I’d still get a meter, though.
Can you compare those numbers against what the DRAC reports itself? The R610 has a real time power usage meter on the web UI of the DRAC.
Just curious how accurate that built in monitor would be. We ran a bunch of benchmarks on R610s last year and originally tried to use PDUs to monitor them but the PDUs were not sensitive enough.
Also if you over spec your power supplies by using 700W when your only peaking at 350W or so is a good way to kill PSU efficiency.
Nate, agreed on the PSU efficiency. The other option is 502 Watt PSUs which might be fine. If you put more drives in a host it’ll drive up the draw, though.
The iDRAC power monitoring interface is very close, actually. It appears to round the PSU draw to the nearest tenth of an ampere, and it doesn’t show current input voltage which makes it hard to check.
I replicated my test above to see what the iDRAC says. PS 1 Current = 0.8 A, PS 2 Current = 0.7 A. I went and physically measured the voltage, which is 202.6 Volts right now. So, (0.8 A + 0.7 A) * 202.6 V = 303.9 Watts.
The iDRAC system board reading shows 329 Watts. Close enough! 🙂 I think they’re either making an assumption about voltage, their measurement apparatus isn’t as sensitive as a Fluke 322, or they are doing some excessive rounding somewhere. Again, still pretty close.
I’m going to correct myself with the iDRAC test — at idle it shows that I’m drawing 0.5 Amps on each PSU. That isn’t correct. At idle I’m drawing 0.35 Amps on each, according to the external meter. I think their monitoring isn’t very sensitive, leading to some inaccuracy (in this case a 30% error).
interesting, good to know.
I talked with a rep from servertech recently about their new POPS (per outlet monitoring) PDUs, and they claimed some pretty good accuracy,
http://www.servertech.com/uploads/datasheets/0000/0076/CWG-16V_C13-C19_.pdf
“Monitor Current Load (A), Voltage (V), Power (W), Apparent Power (VA), Crest Factor and Power Factor per outlet at an accuracy of ±2%.”
He went into a little detail on how they achieve the accuracy but I don’t remember specifics.
Bob, on the PSU efficiency angle: did the server load-balance evenly between 2 PSU’s at peak load, or did it favor one (thereby increasing efficiency)?
Thanks for the info!
When your servers have a 2.3:1 ratio of power consumption loaded vs. unloaded and you pack them tighter than the rating, you run the risk of a power circuit break when your services get really popular and a 22 amp average draw rack suddenly spikes to 50.6.
Alex – it was close to even, which may decrease efficiency but is nice because it spreads the load among both circuits the server is attached to. That means in our data center the left PSU can be on the left PDU, the right PSU on the right PDU, and we don’t have to do funky things about juggling them around to spread the load out.
Rob – that’s why you size for peaks, run at 208 volts to decrease the draw, etc.
Very nice, Bob!
Hi, This article provide a very good example of how an idle server can consume power. Jut one thing to point out, when you computes the power consumption for AC, it is not simply current*voltage, as we still have a current/voltage phase difference to consider (power factor). so it is better to be measured using a power meter, as I remembered some of these PC benchmark sites mentioned such devices. Like referenced here:
http://www.pcworld.com/businesscenter/article/150321/power_plug_meters_help_save_money_energy.html