Computing time will ALWAYS be cheaper than human labor.
back
1 comments
I'm not sure I agree with this. When operating at a very large scale, like Google, eeking out tiny gains in efficiency can yield huge cost savings. A 1% saving of CPU time on a 100K core cluster means 1,000 cores can be eliminated from the system. Figure $50 per core, and you've just saved $50,000 in base cost, plus a ton on cooling + power. Meaning, if it takes less than 3 dev-months, it is absolutely worth the developer time to optimize.
This is absolutely a true statement, and its wrong :-) Until the proposals for energy proportional computing [1] bear fruit there is a lot of 'sunk cost' going into the infrastructure. So finding useful work to do with that sunk cost is a net win.