Saturday 29 August 2009

Supercomputer in the firing line over carbon footprint

It's impossible to ignore the irony that the Met Office supercomputer dedicated to modelling climate change has an enormous carbon footprint. But critics just miss the point
In the past I've written about the internet's energy footprint, and raised the question of whether we should consider ways to drastically reduce the power consumed by data centres as our lives go increasingly online.
So it was with more than a little interest that I read this story about the Met Office's weather-predicting supercomputer - and how, ironically, it has a pretty big carbon footprint.
According to a study by the Department of Communities looking into the footprints of public buildings around the UK, the £33m IBM cluster produces up to 75% of the carbon emissions from the Met Office's HQ in Exeter. That means it's responsible for a good deal of their annual output of 12,000 tonnes of carbon dioxide - one of the highest totals in the country.
It's an easy headline, of course - and probably feels particularly sweet for those who revel in poking at climate change scientists (and they are certainly a vocal group). But the delight in this irony is also driven, in part, by the misplaced assumption that a climate scientist would think that all carbon emissions are bad.
Are all CO2 emissions equal? Well I'm certainly happier to see carbon being spent solving the world's biggest problems than ferrying people around the globe for business meetings they could do through teleconferencing.
And, of course, it's all relative in any case. Supercomputers in the past used vast amounts of energy to run, and were extremely primitive by comparison to today's beasts - on a visit to the Museum of Computer History earlier this year, I saw an old IBM machine that used the same amount of power as a small town but had just a few KB of memory. The Met Office machine hopes to be able to run at a Petaflop soon - that's 1 thousand trillion (1,000,000,000,000,000) calculations every second.
The truth is that while the public rarely thinks about supercomputers - except, perhaps, when there's a chess game at stake - these machines do amazing work that is impossible to replicate elsewhere. These are the machines that fold proteins, that crunch data from particle accelerators; they are the sort of machines that could help cure cancer - or, in the case of the Met's cluster, save thousands of lives by accurately modelling a climate-related disaster. It's worth thinking about.