Hi,
On Wed, Dec 14 at 10:33, James Courtier-Dutton wrote:
> Hi,
>
> I have an aim to monitor as precisely as possible the power usages of
> a number of servers.
Like:
# ioline-summary
Data Out power: Bus Voltage 13.38 V
Data Out power: Current 0.256 A
Data Out power: Power 3.43 W
External power outlet 0: Bus Voltage 12.87 V
External power outlet 0: Current 0.002 A
External power outlet 0: Power 0.02 W
External power outlet 1: Bus Voltage 12.86 V
External power outlet 1: Current -0.001 A
External power outlet 1: Power -0.02 W
Port A power: Bus Voltage 12.86 V
Port A power: Current -0.069 A
Port A power: Power -0.89 W
Port C power: Bus Voltage 12.84 V
Port C power: Current 0.001 A
Port C power: Power 0.02 W
Sensor power: Bus Voltage 12.87 V
Sensor power: Current -0.001 A
Sensor power: Power -0.02 W
Sorry to tease, that's on a piece of non-PC equipment and uber expensive.
Oh and Kelly wrote the interface so users can get graphs against time and
other parameters such as temperature.
Standard PC hardware doesn't normally allow you to monitor the current
drawn from a supply, only the voltage, hence no way to determine the
power consumption. Voltage was easy to measure so was thrown in to keep
customers amused, but current would have cost a few pennies so wasn't.
A few specialist suppliers and high end manufacturers do include current
monitoring hardware. Look for ESA compatible power supplies for example.
> I cannot seem to find any tools that monitor things like:
> 1) Total power consumbed as a monotonic kWh value.
Unless you've got one of those expensive monitorable power supplies your
best bet is going to be a CurrentCost appliance monitor or similar.
> 2) Power broken down by device within the system.
Without the sensors to measure it software is only going to be able to give
you the broadest stroke estimate. Again per device current sensors is
rather specialist.
> 3) Power broken down by process within the system.
Even with hardware sensors ascribing power consumption to individual
processes would be difficult. I guess you could look at a processes I/O
stats then do something like 10% of disk I/O equals 10% of disk hardware
power consumption.
I think a lot of the monitoring software out there is smoke and mirrors
on top of guesswork.
--
Bob Dunlop
--
Please post to: Hampshire@mailman.lug.org.uk
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--------------------------------------------------------------
