I was thinking today about the cost of speeding up the time to get, say 5 blocks generated if one was impatient for a purchase to be confirmed.
So, I rented a server at voxel.net, which claims to cost one third what amazon does, and I ran bitcoin for a bit to see what it could do there, per core.
Results: about 1000 kh/s per core. (Note, I expected double or triple based on their marketing. Some tuning could probably be done here.)
Incidentally, this implies cost per minted BTC of (at current difficulty of 354): $.1/hr for roughly 17.5 days = $42 per 50 BC = $.84 per BTC. Ouch!
But, let's imagine I want to double up the speed of the network to get my transaction going quickly. I would then need approximately 2.5 million kh/s, or about 2,500 servers. This would get me roughly one block per 10 minutes. Combined with the rest of the network, we'd be double-timing blocks, or getting one out every five minutes. Of course, I'd only need to do this for about 25 minutes to get my five blocks.
So, 2,500 servers * .42 hours = 1,041 hours * $.10 = $104.10. Less the current value of the 250 BTC at $.06 = -$15: Net cost is $90 to get the transaction sped up.
If you wanted the confirmation in 10 minutes, you'd need four blocks plus the network's blocks, so you'd need roughly 10,000 servers, but of course, they'd take less time:
10,000 server * 1/6 hour = 1,666 hours = $166 - $15 = $151.
Someone less lazy than me could put this into a closed-form equation, with a confidence interval, rate and number of blocks specified, I'm sure.
I find this really interesting: right now someone could profitably offer a 'turbo' button for roughly double these numbers, so about $200 to $300 per 5-block generation. This would also not massively impact difficulty the next block cycle as it only ran briefly.
Thoughts or comments?