Over the next several weeks we will be focusing on “Green Computing Technology” and will be presenting articles from those who are currently working and developing in this area. One of these companies is Greenqloud that we have previously covered on our:” Top 25 European Cloud Computing Rising Stars List“
In the meantime, here is an excellent article written by Eirikur Hrafnsson, CEO & Co-Founder Greenqloud that was originally posted back in October 2010 discussing a Power Usage Effectiveness Scale or PUE.
You can expect future contributions to CloudTweaks from Eirikur Hrafnsson as well as Environmental Manager, Ruth Shortall in the following weeks…
A few years back the Green Grid developed the Power Usage Effectiveness metric or PUE for short to measure a data centers’ effectiveness of getting power to IT equipment. What the PUE tells you in simple terms is how much extra energy you need for each usable KWh for the IT equipment due to the power going into cooling, power distribution loss etc. and it’s a simple formula (in theory):
PUE = Total Facility Power/IT Equipment Power
However the PUE “standard” has been the center of the data center industries’ debate ever since it came out. I won’t go into the details but it has been debated mainly for four reasons.
- The PUE can change dramatically depending on where measurements are made, when they are made and the timespan the measurements are made in.
- Data centers are subtracting factors from their PUE to lower it e.g. district heating.
- PUE was designed for dedicated data centers (preferably full of gear) but being calculated for any kind of “data center”.
- PUE was NOT designed as a metric to compare data centers for business purposes (My PUE is smaller then yours!) but as a metric for data centers to improve their effective use of power.
In other words PUE has loopholes and they are being exploited for marketing purposes. Green Grid has decided to plug most of those holes with an update to the PUE definition and are now calling it PUEx where the x is a value between 0-3 and depends on e.g. where measurements are made. You can read more about the PUEx here .
PUEx is a welcomed update and I hope data centers will soon start updating their PUE’s. I still have a beef to pick with the marketing departments of the data centers though because I don’t see them stopping to use PUE as tool to compare their “green” data center to others. The age of “Green by PUE” has just started, believe me!
This article about “Chicago emerging as a green data center powerhouse” is exactly the sort of “greenwashing” we have been spotting here at Greenqloud. Microsoft has a massive “green” data center in Chicago that they say has a PUE of 1.22, sounds pretty great right? But wait…where does their power come from? 72.8% coal, 22.3% nuclear, 3.8% mixed, 1.1% renewable…still sounds green to you? Me neither and we feel like something has to be done. PUE rightfully does not change with the amount of renewables used. It simply isn’t a metric about true greeness but how well you manage your power and therefor should not be used for green marketing.
However, PUE IS being used for green marketing and realistically I don’t think that will change for a while. We therefor came up with a way to “weigh” the PUE to better see which data centers are truly green in the sense that they indirectly cause the least amount of CO2 to be emitted by their use of dirty or clean energy.
This new metric we call GPUE or Green Power Usage Effectiveness and it does two things for the industry.
- It gives the truly green data centers a boost in the battle with their “greenwashing” counterparts.
- You can very easily calculate the CO2 emissions created for each usable KWh for the IT equipment.
So here’s the definition and the presentation format for GPUE:
GPUE = G x PUEx (for inline comparison of data centers)
or = G @ PUEx (a better display and for co2 emission calculations)
The “G” is the key factor here and it is a simple calculated value:
G = Weighed Sum of energy sources and their lifecycle KG CO2/KWh
G =∑( %EnergySource x ( 1 + weight) )
What’s the + 1 for? It’s there so we when “weigh”(multiply) the PUE we get a number that’s not less then the PUE. This gives us a better feeling of the scale because we are used to the small PUE value range.
The weights are simply taken directly from the “lifecycle CO2/KWh for electricity generation by power source” table above we got from the 2008 Sovacool Study e.g. the weight for unscrubbed coal is 1.050 (kg of CO2/KWh) while hydroelectric river generation has a weight of 0.013. An unknown energy source or a “mix” will get the same as the maximum value which for now is the same as coal.
PUE 1.20, 50/50 Coal/Hydro
G = 0.5*(1+1.050) + 0.5*(1+0.013)
G = 1.531, GPUEx = 1.84 or firstname.lastname@example.org
Kg CO2 per usable KWh = (G-1) x PUEx = 0.64 Kg
It’s very interesting to see the GPUE side by side to the PUE of these “green” data centers in the image below (click for a large version). This table is mostly taken from the Greenpeace report “Make IT Green“. You will notice the percentages don’t quite add up, that’s because these are the local grid suppliers, the rest is what we call a “mix”. Some of those data centers also didn’t have a public PUE (underlined) so for demonstration purposes we gave them a PUE of 1.5 and did the calculations. Greenqloud’s data centers don’t have their PUE yet but from Icelandic experience and the free cooling here we are certain that they will at least be under 1.2 and because we only use 100% renewable energy (geothermal and hydro) for our truly green public cloud our GPUE will roughly be equal our PUEx and that’s the point. Truly green data centers will have a GPUE close to their PUEx and dirty energy data centers won’t, let the debate begin!
Please leave your comments, share on twitter etc. I would love to hear your thoughts on the subject!
Contribution By Eirikur Hrafnsson, CEO & Co-Founder Greenqloud
Latest posts by CloudTweaks (see all)
- Video Conferencing For Effective Customer Engagement - February 27, 2015
- Cloud Infographic – Healthcare Wearables - February 25, 2015
- Mirantis Collaborates With Google on Kubernetes and OpenStack Integration - February 24, 2015