kWhr to kgCO2 conversion rate for the ESB?

10-10-20

A project I'm starting requires me to measure the efficiency in terms of cost and pollution of a server lab used for testing and replication.

The lab has 15 server racks, all full and mainly all powered on. On calculating the consumption of one of these racks I came to a figure of 3420W and a yearly consumption (it's powered on 24hrs by 7days) of 29,877kWhr.

According to a formula I got online from a UK site, 1kWhr = 0.43 of a kilogram of Carbon Dioxide. So in effect, one of these fifteen racks indirectly produces 12847.16 kg's of CO2, or 12.8 tonnes of the stuff per annum!

Can anybody tell me whether the ESB provide a kW to CO2 factor based on their output so that I can have a true figure rather than the UK rate which includes nuclear power (i.e. no CO2)?

maniac101

The official figure for 2006 is 601g CO2 per kWh. This will reduce further as more combined cycle gas plants and wind turbines are added to the system in coming years.
linkie
(Don't forget the air conditioning in the server room!)

10-10-20

Wah... 18 tonnes p/a... per rack.

Thanks for the info!

ircoha

Just musing here:
30,000kwh per rack per year

15 racks
450,000 kwh
at 0.16 euro per kwh= 72,000 euro in power alone for the servers.

Does this stack up: just wondering:)

Nukem

ircoha wrote: »

Just musing here:
30,000kwh per rack per year

15 racks
450,000 kwh
at 0.16 euro per kwh= 72,000 euro in power alone for the servers.

Does this stack up: just wondering:)

A little confused here. 30,000kWh PA?

If thats so the individual rack uses 30,000/(365*24)=3.42w, seems very small for a server power supply. (not an IT expert).

If 3.42w is correct then all well in good., but remember will not be at max demand all day. See http://enterprise.amd.com/Downloads/svrpwrusecompletefinal.pdf

The only way to get a fair measurement of this is to throw a clamp on CTs with an energy meter locally for a few days and work it out through that. Or if you have an amp meter to check the amps every few hours, that will give you statistical data and a fair reflection fo consumption.

10-10-20

Nukem - 30,000kW/a = 3.42kW per hour (not watts!).
And, the figures are based on idle power and not based on the servers running full tilt! Add another 1/3 for fully loaded - but this event doesn't occur as it's a lab envronment.

I was going to verify these calculations with a clamp mater but I had always assumed that a clamp meter would measure the load on a standard mains cable, but of course they don't - you've got to expose the two feeds and measure from there. So, I'm going to use a plug-in meter from Maplin...

Irchoa - we get the ESB's 'enterprise' rate of €0.10 per kW, and not all of the racks are populated or fully powered up.

So far, I've calculated 10 racks with the following results:

Watts (Calc): 20,152 W
Tonnes CO² p/a: 106.17
ESB Cost €: €17,665

I'll feedback once I know more!

10-10-20

Proxy here bit the dust and I double posted...

Nukem

Apologies i misinterpreted what you had written.

10-10-20

Nukem wrote: »

Apologies i misinterpreted what you had written.

No worries - input is welcomed!

OTK

Have you included the power requirements for the UPS system(s)? Have you included power requirements for air conditioning?

10-10-20

Ah, OK, let me cover that off now.

UPS's are not implemented in the lab except on a couple of systems (backup batteries for SAN hard drive arrays) so they only consume power for a minimal amount of time (2-3 mins). I'm not going to factor them in.

AC is a little harder to factor in electrically as we have a dual AC setup (not redundant - again it's a lab not a working environment) but I'll work it based on the fact that the heat being generated by the servers is equal to the wattage input (ok, we do have fans and hard drives in there and the factor is more like 90%) and the cooling should match that consumption.
I will be able to measure what the AC units pull in terms of amps later on, but I'm just concentrating on the physical servers at the moment.

Do you happen to know how to convert BTU to cooling power in watts, or is that dependant on the manufacturer of the AC unit?

OTK

I don't know how to measure power requirements for air conditioning systems. They go on and off during the day and use more power in the summer.

I doubt you can assume that 90-100% of heat generated by the servers is equal to the power use by the a/c. Is a lot of heat not naturally dissipated?

10-10-20

You're correct in saying that about some AC systems, but we have a stable supply of heat from the servers, winter or summer, so the rate should be roughy the same.
The heat generation for electronic devices is considered to be above 90% of the watt input. With hard drives, some of this power is consumed in the energy required to run the motor, but it's still 75% or similar.

Dissapation of heat isn't correct, the heat is still going somewhere - be that into the office next door or above. It still has to be removed through AC systems.

I'm going to ask the facilities guys to advise on the details of the above, but I believe it to be correct.

Capt'n Midnight

If all the servers were on UPS then you could read the power usage from their software. Clamp meter might work at the distribution board, as long as the sparks have bothered to label the neutrals properly (usual warnings about getting a professional to do this).

Depending on what they are doing you may be able to enable power saving / Wakeup On Lan or set the BIOS to wake up in the morning. If they are number crunching 24/7 not much you can do.

Can you use the heat elsewhere as an input to another AC unit or duct the air to a green house or something.

maniac101

10-10-20 wrote: »

The heat generation for electronic devices is considered to be above 90% of the watt input. With hard drives, some of this power is consumed in the energy required to run the motor, but it's still 75% or similar.

This looks a little high to me. Each piece of electronic equipment in the room should have a rated heat dissipation value documented in its technical specifications. This would be a good starting point for calculating the additional air conditioning load. The power used by the AC to remove the heat from the space will depend on the AC design, the type of cooling used and the efficiency of the chillers (typically ca 80%).

I was thinking that you could take a more controversial approach of also looking at the CO2 emissions avoided by the presence of the servers, through greater productivity, data security and integrity etc. Of course these avoided emissions would be a lot harder to quantify, but the figures could be massaged to be a multiple of the actual emissions of the servers themselves! Depends on the ultimate purpose of your report I suppose.

10-10-20

I just checked this figure - for electronic devices with no internal moving parts (fans, hard drives) the input power to heat output factor is 100%.
For fans alone, the ratio is more like 40% with some of that consumed in the creation of noise, vibration and the power to spin the fan blades.

On the server manufacturers site the efficiency (power in to heat out) is taken to be 100% of the consumed power even when hard drives make up 95% of the consumed power.
For example, if I configure an server with two hard drives, the specifications tool outputs the following information:
System Heat/Power 309.6 W/h

I can also get this in BTU/h.

Mailman

Hi,
digging up old post here.

I've just looked at my ESB bills and took the meter readings off them.
between 10/1/7 and 11/1/8 I used 4259kWh for both day and night tariff ESB.
601g per kWh was the official figure for last year according to link earlier in post.
I multiply 601 by 4259 and get 2559659grams total.
which is 2560Kg.

Can I really have consumed enough electricity to have released 2.56 tonnes of carbon in the last year?

That seems like an awful lot of CO2 for just little 'ol me.

I did look at my electricity usage halfway through the year and made some changes which should hopefully show in my bills/meter readings for the coming year.