1650 power consumption.

Ben Russo Ben at umialumni.com
Tue Jan 21 09:44:01 CST 2003


Steve Mickeler wrote:

>Can anyone let me know the average current utilized by a dell 1650 with 2
>power supplies ?
>
>2 x 275W power supplies = 550W
>
>550W / 220V = 2.5 amps.
>
>2.5 amps * 40 servers in a rack = 100 amps
>
>Im guessing (hoping too) that they dont suck 2.5 amps for 7*24
>  
>
I dunno about Dell 1650's, but I can tell you about my experience with 
densely racking
IBM, SUN, CISCO and 3Com equipment.

We bought a little Wattage meter at Home Depot and plugged the servers 
into it in
a lab setting,  we found that a box with 2 SCSI disks,
2 750Hz PIII's, 1GB RAM, dual NIC's drew up to 200W (i.e.  when booting up)

This was on boxes that had power supplies with a max draw of 3A @ 120V
IIRC However once the server was booted and was just running, the boxes 
drew
anywhere from 60-170 W , usually averaging around 75 watts.

So, we had Rack mountable APC power management devices that allowed us
to stagger power up from outage (e.g. if you unpluged the power strip it 
would
apply power to the outlets one at a time, and would stagger them by a 
few minutes)


We found that SUN Sparc, IBM PA-RISC, and all PC hardware and such
all had similar characteristics.

This made sense anyway because we wanted the DISK arrays, and Network
Devices powered on first, then wait 2 minutes, then power on the 
FW/DNS/NTP/DHCP/ROUTERs,
then wait 2 minutes, then power on the NFS/LDAP/Domain Controllers, then 
wait 5 minutes,
then power on the other devices 1 per minute per power strip.

Now, we had two power strips in each cabinet (from seperate Liebert 
power rectifiers, that were
backed by seperate building UPS, that were backed by seperate sets of 
Generators).
And most of our equipment was either fault-tolerant by being redundant
in which case we had the A unit in the A power circuit and the B unit in 
the B power circuit.
Or we had servers in groups of n+1 for redundancy where the servers had 
redundant power
supplies and were "dual-homed"

So for most of our Racks we would never load a single power circuit to 
more than 50%,
because if we had a power phase outage, then the matching circuit would 
instantly get
almost a double load from all the servers with redundant power supplies 
that would now
be getting all their juice from one side.

However, we would calculate the servers *AVG* as a rule-of-thumb, 50% of 
the peak
rated draw listed on the Power Supply, but we had to make sure that the 
power strips
would boot up the servers one at a time.

-Ben.






More information about the Linux-PowerEdge mailing list