Thoughts on British ICT, energy & environment, cloud computing and security from Memset's MD
See updated article which includes network energy, here.
A few weeks ago I was party to an initial ‘getting together’ in BAFTA of some key greenie people from the IT industry (me and Emma Fryer, author of many of Intellect’s thought-leadership documents and responses to government policy in the field of carbon reduction) with leaders from the media and entertainment industry. One of the contentious topics is around the environmental damage of CDs compared with downloads. In this article I will calculate both using common sense and first principals.
One of the most interesting things I heard that day was that the music industry’s carbon footprint (which is not that bad – around 2% of total or some such) does a whopping 25% of their environmental damage due to festivals, and that 40% of that is due to people driving to them!! When I went to Download last month we (2 adults, 2 kids) hired a diesel and drove there because it was cheaper than going by train. Sheer madness! Can’t blame the music industry for the inadequacies and negative-economic-incentives around our national mass transit systems.
The music industry has commissioned research with Oxford university which, it claims, shows that CDs are “not that bad compared with downloads”. Now to me with my technologist hat on, that makes no sense at all. How can sending data around on thin bits of heavily processed million-year-old sea-critter be even close in energy efficiency to shuffling some electrons and photons down wires and fibres?
Before starting, I find it useful to get some real-world perspective on the numbers. We are going to be talking about electrical loads, so here are some examples: a modern ceramic kettle uses 3,000Watts (W), my hair dryer uses 1,500W, a modern TV on standby uses 1W and a 100W filament light bulb uses, wait-for-it… 100W!
The human body uses 100W, and your brain uses 20W. Unlike a computer, though, your brain does not use more energy when you think harder. 😉
Lets assume a 2U, 10 Tera Byte (TB) server (Dell PowerEdge 2950 with 6x2TB disks) with a 100Mega bits per second (Mbps) uplink capacity. For resilience, lets stick three at different locations around the UK and have some redundancy (any one of the three can be offline).
Each of those machines will use about 300Watts (W) when working moderately hard (we benchmarked them ourselves), so stick them in a mediocre PUE 1.5 data centre and we have 450W for one, 1,350W for all three. Networking kit will be a minimal overhead, easily absorbed into the PUE.
To be complete we should include the embedded carbon of the servers too. According to my educated guesstimate of embedded energy, an average server probably uses 1,000 kWh in its manufacture. These are big-ish servers, so lets call it 1,500 kWh. Spread over a 3 year working lifetime:
1,500,000 Wh / ( 3 years * 356 days * 24 hours ) = 58 W effective consumption
Three lots of that becomes 175W, added into our 1,350W becomes 1,525W. As you can see, the power required to run the servers completely dwarfs the power consumed by them during their lifetime.
Rummaging through my iTunes folder, it looks like an album averages roughly 100 Mega Bytes (MB) (7-8MB/song, 12-14 songs). 100 MB at 2 Million bits per second (Mbps) (the average we are all supposed to have by 2012 according to the Digital Britain report) gives us 6 minutes, 40 seconds to download an album.
Assume that we leave the servers on all the time, and that we only use them for 8 hours in every 24, but while we are using them we are maxing 2/3 out (200 Mbps), we get a download capacity of:
(( 200 Mbps / 8 bits ) * 3600 seconds * 8 hours ) = 720 Giga Bytes (GB) per day, or 7,200 albums per day.
Our servers, on continuously (and erring on the side of caution) will use:
1,525 W * 24 hours = 36,600 Watt-hours = 36.6 KiloWatt-hours (KWh)
That gives us an energy used to host and serve the album (including life cycle carbon cost of the kit) of:
( 36,600 / 7,200 ) = 5.1 Wh per download.
By way of comparison, bringing a litre of water (half a kettle) from room temperature to boiling takes 87 Wh. Not a lot to write home about so far!
– What about the PC?
So, what about the home PC? Well in the infamous “Google searches use loads of energy” story journalistic incompetence ruled and they included 15 minutes of high-performance home PC energy usage in the calculation, which totally dwarfed the energy used at the data centre.
If you are curious, Google uses commodity hardware, and tells you how long a search takes. A fair assumption would be that 2-3 servers get used in one search, and that they are using about 100W each. A search for “blah” takes 0.12 secs. Add in their PUE of around 1.2 and you get:
( ( 100 W * 3 servers * 1.2 ) * 0.12 seconds ) = 43.2 Watt-seconds = 0.012 Wh
Your brain used twice that reading this sentence!
Anyway, lets factor in the PC so that the raving-lentilist-luddites don’t dismiss this out of hand: At home, a modern PC + screen uses around 150W and a laptop uses perhaps 80, so lets call it 120W. Lets also assume you are either staring at screen or gone away and not using it while the music downloads (silly scenario, but going for worth-case). That gives us:
120W * ( 400 / 3600 secs ) = 13.5 Wh.
So, our grand total per album download is about 18.6Wh. Enough to boil a quarter of a litre of water (one cup of tea). Assuming a British mix of power generation (0.718g/Wh), that is equivalent to about 13.4g CO2.
So, how do we measure a CD? This is tricky, since there are negatives and positives arguably. On the one hand, the manufacturing process is energy intensive, as is delivery, but on the other this could be viewed as a form of carbon sequestration – just as buying books and keeping them on your shelves is good from a climate change point of view.
Lets assume 1,000 CDs in the back of a transit van, being delivered from EMI’s factory to an HMV outlet. The CDs are probably going to have to travel on average at least 200 miles. A van will do something like 400 miles on 40 litres, so 20 litres for our 1000-CD journey.
Petrol is 0.73g/cm3 and 90% of it by weight is carbon atoms. Thus, it is 0.66g/cm3 carbon, or 660 g / litre, which when you add on two oxygen atoms to each carbon atom becomes about 2,000 g / litre.
Therefore, our deliver carbon is:
20 litres * 2,000 g = 40,000g / 1000 CDs = 40 g CO2 per CD.
I’m being generous and ignoring the carbon-cost of manufacturing and maintaining the van, but lets keep it optimistic.
– Packaging & the disk itself
According to the Music Industry’s own publication “Julie’s Bicycle: Reducing the carbon emissions of CD packaging” a plastic CD box’s manufacture causes around 350g CO2, a card wallet would cause about 20g CO2, and the CD itself is around 100g CO2.
Lets by really optimistic and only look at a CD in a card wallet; that is around 120g CO2, which becomes 160g CO2 when delivered to your local HMV store. Oh dear.
Downloading an album: ~14g CO2
(including the time to power the PC while downloading)
Creating & transporting a CD: ~160g CO2
(>500g CO2 if its in a plastic box)
However, what the music industry themselves seem to have failed to notice is that, just with books, all the CDs on your shelves are mostly carbon, and if you don’t throw them away that is a great form of carbon sequestration*. Personally, though, I’d rather download music and have a pot plant where my CD stack currently lives, and that is definitely more environmentally friendly.
* As pointed out by Doug (see below) CDs are of course not a form of carbon sequestration comparable with books since the carbon in plastic comes from fossil fuels, which are quite happily sequestered away until we go and dig them up!