Ask HN: Which is better. A server farm somewhere hot or somewhere cold?
One is somewhere hot (North Carolina, Apple) [1]
The other is somewhere cold (Sweden, Facebook) [2]
Hot, (assuming more solar energy available) benefits are increased power from a solar plant but the farm requires more cooling. Cold, requires less cooling therefore less energy.
It's interesting [to me at least] because it raises the question of efficiency. i.e. do you build something that requires loads of power - but who cares 'cos we got tonnes of renewable energy... or do you try an require a lot less energy even though hydroelectric is not damaging to the environment the more you produce.
[1] http://arstechnica.com/apple/news/2011/10/apple-building-171-acre-solar-panel-farm-to-power-nc-data-center.ars
[2] http://www.techiespider.com/2011/10/27/facebook-build-server-farm-northern-sweden-arctic-circle/
8 comments
[ 3.5 ms ] story [ 25.6 ms ] threadAgainst that, there are transmission losses. If your data centre in a cold dark place is powered by a solar array in a hot sunny place, you need less energy to spend on cooling, but you have less energy to spend because more energy will be lost in transmission. I don't know how they balance out.
(And if you have enough clients in the hot place who want low latency, it makes sense to put the data centre in the hot place even if it's less energy efficient.)
eg Latency vs cost.
Being based Antarctica can be very cheap and fast. But there's poor connections.
It's like a PHP vs C++ argument. While C++ is secure, efficient and takes longer to implement while PHP is cheaper, faster to implement for MVP.
tl;dr: Doing science with servers in bulk? Better in cold. Doing business? Better more closer to customers that tend to be in warm places.
December 2010 data for Ashville, NC:
Source: http://www.nws.noaa.gov/climate/index.php?wfo=gspData for Greenville Spartenburg for December 2010 is similar but with less snowfall locally.
What you want is access to cheap electricity and cheap cooling. Whatever minimises TCO is where you want to be.
I'd go on to say that the energy coefficient between 35 Deg N and 55 Deg N probably isn't enough to offset the equation significantly.
As such I think that removing heat more efficiently via naturally cold climates would have a greater impact on bottom line. Check out wikileaks cave-based data centers in old Swedish nuclear bunkers. http://bitshare.tumblr.com/post/2383169988/the-wikileaks-ser... (Sweden's bandwidth to the rest of the world also blows ours away.)