CloudCampI went to a CloudCamp in London back in early July that was all a bit weirder than usual – quite a bit of swearing from the speakers, some very passionate lightning presentations and a generally more militant and chaotic feel to the whole event – which made it all the more enjoyable. There was a mix of the strange, the vaguely self-promotional and the fascinating, followed by beers and pizza. What’s not to like!

I arrived a little late due to “just one more email” syndrome and when I walked in Francine Bennett from MastodonC was already presenting some data on the relative carbon footprints of the various Amazon Web Services (AWS) regions. She wasn’t picking on AWS specifically – but using the market leader as a benchmark to get her point across. I’ve found a similar presentation from Francine at a previous event here.

Her presentation reminded me of the points made by Dr Chris Tuppen in a white paper published on the Smart421 web site some time ago (there’s a link to it here)…

  • The data centre industry strives for the mythical PUE (Power Usage Effectiveness) of 1.0, but efficiency does not equate to being green. Being efficient in use of cooling, lighting etc is obviously a good thing, but actually the key issue is…
  • The power source – the dominant factor on the carbon footprint of a data centre is its power source. Kinda obvious really, but the variability in the carbon emissions per kWh around the world is staggering – have a look at the following data…the “bad boys” are typically coal-generation based, and the “good boys” are hydro/thermal or nuclear-based. Greenland isn’t in my data set below, but that would be even better.

Co2 emissions by country 2009

  • So what this means is that if your data centre runs on power from France, it already has nearly 9 times lower CO2 emissions than if it were running in Ireland. So working really hard to tweak your PUE down from 1.2 to 1.1 seems kinda secondary in this context…
  • Another key factor is the climate/environment your data centre is in – i.e. is it in a hot or a cold place. Hot = bad as all the cooling required just drives up emissions, and would of course be reflected in your PUE rating. So Ireland wins here as it’s not that warm, and Greenland really wins.
  • Applying this to an AWS context – what this means is that you can immediately get a 43% emissions reduction by moving your AWS workload from US-East in Virginia (the cheapest region to use, but it’s hot there and the local power sources are dirtier) to the EU Region in Ireland. [Data sourced from here]. Given that moving AWS workloads between regions is pretty trivial to achieve, for scenarios where bandwidth/latency considerations are not a barrier this is an absolute no-brainer to do. This is where IaaS/cloud computing really wins out – the portability of workloads mainly due to the pricing model. Of course – all the Smart421 AWS deployments tend to be in the EU region anyway as it is the local region for our customer base :)
  • But if you really want to get emissions down, port your IT workload to GreenQloud in Greenland who run a 100% carbon neutral data centre facility – this is only possible because they rely on carbon neutral power sources (geothermal and hydro), and helpfully it’s a bit chillier there too. The reality today though is that latency (an assumption – I must measure this at some point…) and the lack of geographically-separated data centres (i.e. the multi-AZ model in the AWS world) makes this not really fit for enterprise use in my view. They run AWS-like APIs for the basic AWS services of compute and storage but of course you don’t get the richness of the rest of the AWS platform.

So – coming back to my opening question – “How green is your cloud and what can you do about it?”. Answer – probably not very depending on where it is, so think about where it is and move it.