linux energy usage
russ
is there a aimple way in software to monitor the energy usage of a
linux machine? can one monitor fan voltage, cpu/disk voltage and
attempt to try and infer overall computer energy use?
thanks in advance
russ
John E. Davis
> is there a aimple way in software to monitor the energy usage of a
> linux machine? can one monitor fan voltage, cpu/disk voltage and
> attempt to try and infer overall computer energy use?
For overall energy usage you will need to use something like
this:
<http://www.thinkgeek.com/gadgets/travelpower/7657/>
--John
Soruk
> Hi
>
> is there a aimple way in software to monitor the energy usage of a
> linux machine? can one monitor fan voltage, cpu/disk voltage and
> attempt to try and infer overall computer energy use?
There's nothing in software unless you're using a laptop when the ACPI
interface can tell you how much power (or the current, depending on your
ACPI) is being drawn from your battery.
The simplest way is to use an adapter like this one:
http://www.maplin.co.uk/Module.aspx?ModuleNo=38343
--
-- Michael "Soruk" McConnell Eridani Star System
MailStripper - http://www.MailStripper.eu/ - SMTP spam filter
Second Number - http://secondnumber.matrixnetwork.co.uk/
Jean-David Beyer
> Hi
>
> is there a [s]imple way in software to monitor the energy usage of a
> to try and infer overall computer energy use?
The way I do it, since I do not really care much, is to look at the
indicator lights on my UPS. If you go here, and enlarge the picture, you can
see the lights on the left side of the unit (as you face it).
If they are all off, you are using less than 17% of the power it can supply.
If the bottom one is on, you are using between 17% and 33%
If the bottom two are on, you are using between 33% and 50%
If the bottom three are on, you are using between 50% and 67%
If the bottom four are on, you are using between 67% and 85%
If they are all on, you are using over 85%. Since my unit is rated at 2200
VA, I can easily figure it out.
The lights on the other side are charge in the battery, but they are easier
to remember: 20%, 40%, 60%, 80% and 100%.
>
> thanks in advance
>
You could also get a power meter, as has been suggested by another poster.
That one would not work for me because my UPS uses a 20Amp plug, and that
meter only accepts a 15Amp plug.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 20:05:02 up 4 days, 5:49, 4 users, load average: 4.45, 4.38, 4.23
russ
> russ wrote:
> > Hi
>
> > is there a [s]imple way in software to monitor the energy usage of a
> > linux machine? can one monitor fan voltage, cpu/disk voltage and attempt
> > to try and infer overall computer energy use?
>
> The way I do it, since I do not really care much, is to look at the
> indicator lights on my UPS. If you go here, and enlarge the picture, you can
> see the lights on the left side of the unit (as you face it).
> If they are all off, you are using less than 17% of the power it can supply.
> If the bottom one is on, you are using between 17% and 33%
> If the bottom two are on, you are using between 33% and 50%
> If the bottom three are on, you are using between 50% and 67%
> If the bottom four are on, you are using between 67% and 85%
> If they are all on, you are using over 85%. Since my unit is rated at 2200
> VA, I can easily figure it out.
>
> The lights on the other side are charge in the battery, but they are easier
> to remember: 20%, 40%, 60%, 80% and 100%.
>
> > thanks in advance
>
> You could also get a power meter, as has been suggested by another poster.
> That one would not work for me because my UPS uses a 20Amp plug, and that
> meter only accepts a 15Amp plug.
thanks all for the above. it largely confirms what i thought, that no
you can't do it in software. i was hoping that i could use cpu usage,
fan, cpu temperature or other proxies to estimate energy usage. the
plug in monitor is fine for my computer but I want to monitor several
hundred to see the energy they are putting into a building (i am an
energy rather than computer geek). is it possible to find the rail
voltage in software? is the fan a good proxy for energy use and is it
controlled by the OS?
russ
TheLostPacket
most BIOSes have rail voltage/fan monitors, I'm pretty sure there
should be some way (eg on a Windows box, with MBM) you can monitor the
output of that in Linux. Don't ask me what though, 'cos I don't know.
Hardware wise, I think someone already pointed a link to an inline
power meter which does realtime monitoring of input power, voltage,
and the total KWh used while the thing's had power going through it.
Very nifty device, and yes it does give surprising results (like, a
VIA C3 Nehemiah 1.0GHz on an Epia M board, with an 80GB notebook
drive, no fans, 22" 65W widescreen and Sky Digital box giving change
out of 100W!)
TLP
Maxwell Lol
> Hi
>
> is there a aimple way in software to monitor the energy usage of a
> linux machine? can one monitor fan voltage, cpu/disk voltage and
> attempt to try and infer overall computer energy use?
Buy a device called Kill-A-Watt, and plug your computer into it.
Hactar
russ <rlay...@hotmail.com> wrote:
>
> thanks all for the above. it largely confirms what i thought, that no
> you can't do it in software. i was hoping that i could use cpu usage,
> fan, cpu temperature or other proxies to estimate energy usage. the
> plug in monitor is fine for my computer but I want to monitor several
> hundred to see the energy they are putting into a building (i am an
> energy rather than computer geek). is it possible to find the rail
> voltage in software?
I believe you can get apcaccess to ask apcupsd to ask the UPS what the
line voltage is, but I don't know any way of asking the computer
directly.
--
-eben QebWe...@vTerYizUonI.nOetP royalty.mine.nu:81
This message was created using recycled electrons.
Jean-David Beyer
> In article <4070fdfa-5659-4cd1...@k36g2000pri.googlegroups.com>,
> russ <rlay...@hotmail.com> wrote:
>> thanks all for the above. it largely confirms what i thought, that no
>> you can't do it in software. i was hoping that i could use cpu usage,
>> fan, cpu temperature or other proxies to estimate energy usage. the
>> plug in monitor is fine for my computer but I want to monitor several
>> hundred to see the energy they are putting into a building (i am an
>> energy rather than computer geek). is it possible to find the rail
>> voltage in software?
>
> I believe you can get apcaccess to ask apcupsd to ask the UPS what the
> line voltage is, but I don't know any way of asking the computer
> directly.
>
(energy).
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 10:15:01 up 4 days, 19:59, 4 users, load average: 4.33, 4.35, 4.34
Grant Edwards
>> I believe you can get apcaccess to ask apcupsd to ask the UPS
>> what the line voltage is, but I don't know any way of asking
>> the computer directly.
>>
> True, but you cannot ask it the current, so you cannot computer the power
> (energy).
And if you knew the voltage and the current, you'd still need
to know the power factor (it's based on the phase difference
between voltage and current).
--
Grant Edwards grante Yow! Is it NOUVELLE
at CUISINE when 3 olives are
visi.com struggling with a scallop
in a plate of SAUCE MORNAY?
Jean-David Beyer
> On 2008-11-11, Jean-David Beyer <jeand...@verizon.net> wrote:
>
>>> I believe you can get apcaccess to ask apcupsd to ask the UPS
>>> what the line voltage is, but I don't know any way of asking
>>> the computer directly.
>>>
>> True, but you cannot ask it the current, so you cannot computer the power
>> (energy).
>
> And if you knew the voltage and the current, you'd still need
> to know the power factor (it's based on the phase difference
> between voltage and current).
>
supply, and if that was all there was to it, it would have a horrible power
factor, if the term would even be properly used when the current is not
sinusoidal.
But this power supply contains power factor correction that gets the power
factor up, from whatever it would otherwise be, to at least 0.95.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 15:05:01 up 5 days, 49 min, 4 users, load average: 4.54, 4.64, 4.61
Grant Edwards
>
>> And if you knew the voltage and the current, you'd still need
>> to know the power factor (it's based on the phase difference
>> between voltage and current).
>
> That brings up a question. My power supply is a fancy
> switching power supply, and if that was all there was to it,
> it would have a horrible power factor, if the term would even
> be properly used when the current is not sinusoidal.
>
> But this power supply contains power factor correction that
> gets the power factor up, from whatever it would otherwise be,
> to at least 0.95.
It used to be common to see PFs as low as 0.70, but I think
most recent PC power supplies that aren't junk have power
factors of at least 0.95, and some are hitting 0.99+.
If the question was "how does power factor correction work?",
then I'd have to plead ignorance (even after a couple of google
searches).
--
Grant Edwards grante Yow! If I pull this SWITCH
at I'll be RITA HAYWORTH!!
visi.com Or a SCIENTOLOGIST!
Unruh
>Grant Edwards wrote:
>> On 2008-11-11, Jean-David Beyer <jeand...@verizon.net> wrote:
>>
>>>> I believe you can get apcaccess to ask apcupsd to ask the UPS
>>>> what the line voltage is, but I don't know any way of asking
>>>> the computer directly.
>>>>
>>> True, but you cannot ask it the current, so you cannot computer the power
>>> (energy).
>>
>> And if you knew the voltage and the current, you'd still need
>> to know the power factor (it's based on the phase difference
>> between voltage and current).
>>
>That brings up a question. My power supply is a fancy switching power
>supply, and if that was all there was to it, it would have a horrible power
>factor, if the term would even be properly used when the current is not
>sinusoidal.
The input to the system is sinusoidal 60Hz power. So it is the phase btween
the 120V input voltage and the input current that defines the power factor.
The ouput of most supplies is DC which always has a power factor of 1.
Now internally it is a high frequency mess.
Jean-David Beyer
> Jean-David Beyer <jeand...@verizon.net> writes:
>
>> Grant Edwards wrote:
>>> On 2008-11-11, Jean-David Beyer <jeand...@verizon.net> wrote:
>>>
>>>>> I believe you can get apcaccess to ask apcupsd to ask the UPS
>>>>> what the line voltage is, but I don't know any way of asking
>>>>> the computer directly.
>>>>>
>>>> True, but you cannot ask it the current, so you cannot computer the power
>>>> (energy).
>>> And if you knew the voltage and the current, you'd still need
>>> to know the power factor (it's based on the phase difference
>>> between voltage and current).
>>>
>> That brings up a question. My power supply is a fancy switching power
>> supply, and if that was all there was to it, it would have a horrible power
>> factor, if the term would even be properly used when the current is not
>> sinusoidal.
>
> the 120V input voltage and the input current that defines the power factor.
Of course that is the definition. But where a power supply receives the
sinusoidal voltage but its current demand is non-sinusoidal, can the idea of
phase or power factor even have meaning? It is easy to measure the phase
difference between a power source and a load comprised of linear passive
elements (e.g., resistors, capacitors, inductors, some motors). But when the
load draws very spiky currents near the peaks of the sinusoidal voltage, you
cannot really say what the power factor is. If you stuck a differential
phase meter in there, the result would depend more on the design of the
phase meter than the load as seen by the power company.
> The ouput of most supplies is DC which always has a power factor of 1.
The concept of power factor with DC loads does not really exist. What power
factor tries to do is figure out what the equivalent resistance is to the
load. Because the power company has to have the generation and distribution
capacity to match the volt-amperes drawn by the load, but they can only bill
for the in-phase component. Hence they want users to have as high a power
factor as possible. For residential users, this is too much trouble, but for
industrial users, they can get a surcharge if the loads are too reactive.
>
> Now internally it is a high frequency mess.
>
>
>> But this power supply contains power factor correction that gets the power
>> factor up, from whatever it would otherwise be, to at least 0.95.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 17:35:01 up 5 days, 3:19, 4 users, load average: 4.20, 4.29, 4.35
John Hasler
> Because the power company has to have the generation and distribution
> capacity to match the volt-amperes drawn by the load, but they can only
> bill for the in-phase component.
They must have _distribution_ capacity to match the volt-amperes drawn by
the load (at least back to their own power-factor-correcting equipment, but
then they must pay for that so it should be considered distribution
equipment). Their generators need only supply the real component of the
power.
--
John Hasler
jo...@dhh.gt.org
Dancing Horse Hill
Elmwood, WI USA
Grant Edwards
> Jean-David Beyer writes:
>> Because the power company has to have the generation and distribution
>> capacity to match the volt-amperes drawn by the load, but they can only
>> bill for the in-phase component.
That's not true (at least not in the US). Many industrial
users get billed based on Volt-Amps or on a combination of real
power and kVolt-Amp-Hours. IOW, they pay more per unit of real
energy as their power factor drops. That's why there's a market
for people who manufacture power-factor correction gear for
large industrial users. AFAIK, residential customers in the US
are only billed for real power regardless of their power factor
(residential customer's power factors are usually higher than
those of industrual customers).
> They must have _distribution_ capacity to match the
> volt-amperes drawn by the load (at least back to their own
> power-factor-correcting equipment, but then they must pay for
> that so it should be considered distribution equipment).
> Their generators need only supply the real component of the
> power.
--
Grant Edwards grante Yow! HUGH BEAUMONT died
at in 1982!!
visi.com
russ
> Hactar wrote:
> > In article <4070fdfa-5659-4cd1-bdb4-ee86437d7...@k36g2000pri.googlegroups.com>,
> > russ <rlaybe...@hotmail.com> wrote:
> >> thanks all for the above. it largely confirms what i thought, that no
> >> you can't do it in software. i was hoping that i could use cpu usage,
> >> fan, cpu temperature or other proxies to estimate energy usage. the
> >> plug in monitor is fine for my computer but I want to monitor several
> >> hundred to see the energy they are putting into a building (i am an
> >> energy rather than computer geek). is it possible to find the rail
> >> voltage in software?
>
> > I believe you can get apcaccess to ask apcupsd to ask the UPS what the
> > line voltage is, but I don't know any way of asking the computer
> > directly.
>
> True, but you cannot ask it the current, so you cannot computer the power
> (energy).
i was also hoping (maybe or maybe not naively!) that knowledge of the
real-time voltage out of the transformer would be indicative of the
current and that this would be specific to a transformer.
Dan Mills
> They must have _distribution_ capacity to match the volt-amperes drawn
> by the load (at least back to their own power-factor-correcting
> equipment, but then they must pay for that so it should be considered
> distribution equipment). Their generators need only supply the real
> component of the power.
Actually the generation gear must supply both, as it is the reactive
power that sets the system voltage.
Grid operators do not (in the first instance) think in terms of voltage
and current, but in terms of MW & MVAr, with the subtle issue that due to
the presence of downstream automatic tap changers, lowering the grid
voltage will increase the current demand and thus IIR losses, combine
this with things like switch mode supplies and sync induction motors
(which are constant power) and you can end up with a system that is
dynamically unstable if care is not taken.
The key thing to realise is that the alternators are synchronous machines
and that (in steady state) all that changes as you increase the prime
mover power is the angle between the rotor pole piece and the rotating
field in the stator.
Basically the torque from the prime mover is trying to drag the rotor
ahead of the field and the field is trying to drag the rotor back into
alignment (By implication, if you supply too much torque for the demand,
the network frequency will creep up).
If you draw the rotor angle as a phasor, then the component at 90 degrees
to the field represents real power, and the component aligned with the
field, reactive power. The exciter field determines the size of the
reactive power vector and the system voltage while the shaft torque
determines system current (and its phase relationship to the system
voltage - referred to by whether a generator is capacitive or inductive).
As far as the original problem goes: Stick the PC in a calorimeter and
measure the temperature rise!
Regards, Dan.
Grant Edwards
> As far as the original problem goes: Stick the PC in a
> calorimeter and measure the temperature rise!
But that wouldn't measure the power that was used to do
computations, only the wasted power!
Honest to dog, in another newsgroup somebody claiming to be
educated in electronics took that position and argued it for
quite a while (I finally gave up, since mere facts didn't seem
to factor into his opinions). AFAIK, he still thinks that
power used "to do computations" doesn't turn into heat.
I suppose a few of the photons generated by the LEDs might
escape out the window and into space without being absorbed by
anything, so not _quite_ 100% of the input power turns into
heat [but that wasn't what the argument was about].
--
Grant
John Hasler
> AFAIK, he still thinks that power used "to do computations" doesn't turn
> into heat.
That depends on the extent to which the computations are reversible (those
done by ordinary computers aren't).
John Hasler
> They must have _distribution_ capacity to match the volt-amperes drawn by
> the load (at least back to their own power-factor-correcting equipment,
> but then they must pay for that so it should be considered distribution
> equipment). Their generators need only supply the real component of the
> power.
Dan Mills writes:
> Actually the generation gear must supply both, as it is the reactive
> power that sets the system voltage.
The power plant operator cares primarily about real power as that is what
determines how much fuel he must buy. The distribution system operator
cares a great deal about volt-amps because that is what determines how much
wire, iron, and insulation he has to buy.
Andrew Halliwell
> thanks all for the above. it largely confirms what i thought, that no
> you can't do it in software. i was hoping that i could use cpu usage,
> fan, cpu temperature or other proxies to estimate energy usage.
Nah, voltage and current are different. You can have a 3V circuit that draws
only a microamp or 2 amps, it all depends on what the circuit is doing and
how it was designed/built.
iF ANYTHING, if a circuit is drawing TOO MUCH current, then the voltage will
drop. Voltage is equivalent to "pressure" while current is the "flow" if you
want to use a fluid metaphor. If the flow rises above the rate the pump
(power supply) can supply, the pressure will drop.
Power is a combination of the two. pressure multiplied by flow.
--
| |What to do if you find yourself stuck in a crack|
| spi...@freenet.co.uk |in the ground beneath a giant boulder, which you|
| |can't move, with no hope of rescue. |
| Andrew Halliwell BSc |Consider how lucky you are that life has been |
| in |good to you so far... |
| Computer Science | -The BOOK, Hitch-hiker's guide to the galaxy.|
russ
> russ <rlaybe...@hotmail.com> wrote:
> > thanks all for the above. it largely confirms what i thought, that no
> > you can't do it in software. i was hoping that i could use cpu usage,
> > fan, cpu temperature or other proxies to estimate energy usage.
>
> Nah, voltage and current are different. You can have a 3V circuit that draws
> only a microamp or 2 amps, it all depends on what the circuit is doing and
> how it was designed/built.
>
> iF ANYTHING, if a circuit is drawing TOO MUCH current, then the voltage will
> drop. Voltage is equivalent to "pressure" while current is the "flow" if you
> want to use a fluid metaphor. If the flow rises above the rate the pump
> (power supply) can supply, the pressure will drop.
>
> Power is a combination of the two. pressure multiplied by flow.
not sure if a discussion about the difference between volts and amps
really answers my point about "i was hoping that i could use cpu
usage, fan, cpu temperature or other proxies to estimate energy
usage". was this meant to be a reply to something else?
Jean-David Beyer
> russ <rlay...@hotmail.com> wrote:
>> thanks all for the above. it largely confirms what i thought, that no
>> you can't do it in software. i was hoping that i could use cpu usage,
>> fan, cpu temperature or other proxies to estimate energy usage.
>
> Nah, voltage and current are different. You can have a 3V circuit that draws
> only a microamp or 2 amps, it all depends on what the circuit is doing and
> how it was designed/built.
That is true. My computer draws measurably more power from the UPS when it
is running four instances of BOINC applications (these are all compute
limited) as when BOINC is not running.
>
> iF ANYTHING, if a circuit is drawing TOO MUCH current, then the voltage will
> drop.
I imagine all computer power supplies are voltage regulated. Because of
this, drawing more current will not result in a voltage drop unless you draw
so much as to overload the power supply.
> Voltage is equivalent to "pressure" while current is the "flow" if you
> want to use a fluid metaphor. If the flow rises above the rate the pump
> (power supply) can supply, the pressure will drop.
>
> Power is a combination of the two. pressure multiplied by flow.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 06:35:01 up 1 day, 15:43, 4 users, load average: 4.36, 4.28, 4.25
Andrew Halliwell
Not current, and without both, you can't know how many watts are being used.
--
| spi...@freenet.co.uk | Windows95 (noun): 32 bit extensions and a |
| | graphical shell for a 16 bit patch to an 8 bit |
| Andrew Halliwell BSc | operating system originally coded for a 4 bit |
| in |microprocessor, written by a 2 bit company, that|
| Computer Science | can't stand 1 bit of competition. |
russ
> russ <rlaybe...@hotmail.com> wrote:
> > On 14 Nov, 08:25, Andrew Halliwell <spi...@ponder.sky.com> wrote:
> >> russ <rlaybe...@hotmail.com> wrote:
> >> > thanks all for the above. it largely confirms what i thought, that no
> >> > you can't do it in software. i was hoping that i could use cpu usage,
> >> > fan, cpu temperature or other proxies to estimate energy usage.
>
> >> Nah, voltage and current are different. You can have a 3V circuit that draws
> >> only a microamp or 2 amps, it all depends on what the circuit is doing and
> >> how it was designed/built.
>
> >> iF ANYTHING, if a circuit is drawing TOO MUCH current, then the voltage will
> >> drop. Voltage is equivalent to "pressure" while current is the "flow" if you
> >> want to use a fluid metaphor. If the flow rises above the rate the pump
> >> (power supply) can supply, the pressure will drop.
>
> >> Power is a combination of the two. pressure multiplied by flow.
>
> > not sure if a discussion about the difference between volts and amps
> > really answers my point about "i was hoping that i could use cpu
> > usage, fan, cpu temperature or other proxies to estimate energy
> > usage". was this meant to be a reply to something else?
>
> My point was, the onboard monitors only measure voltage.
> Not current, and without both, you can't know how many watts are being used.
>
accuracy required. to first order, an estimate based on the time the
computer has been in any of the standard power states
http://en.wikipedia.org/wiki/Advanced_Configuration_and_Power_Interface
should suffice
Dan Mills
> Dan Mills writes:
>> Actually the generation gear must supply both, as it is the reactive
>> power that sets the system voltage.
>
> The power plant operator cares primarily about real power as that is
> what determines how much fuel he must buy. The distribution system
> operator cares a great deal about volt-amps because that is what
> determines how much wire, iron, and insulation he has to buy.
And the network dispatcher cares more then a little about real and
reactive as that is how he keeps from blacking out the east coast.
The stability problem for a big power grid is fascinating by the way.
Obviously the prime mover is sized on real power.
On the computation thing, ordering the magnetic domains does imply a
certain amount of change in entropy, but it is minuscule. **ALMOST** all
the power input comes out as waste heat.
Regards, Dan.