Exascale Computing is the challenge of programming for the very large parallel computers which are likely to come about in the next few years. The name comes from the extrapolation of the floating point power in successive powers of ten, as represented by the top 500 computer installations in the world. 

1 flop = 1 floating point calculation per second.

* exaflop  10^18 flops
* petaflop 10^15 flops 
* teraflop 10^12 flops
* gigaflop 10^9 flops
* megaflop 10^6 flops

The point is that computing at this scale requires thinking hard about the algorithms needed to get the required very large scale parallel operation.

There has been a recent conference on this in Edinburgh, Scotland about this - see http://www.easc2013.org.uk/

----
See ProgrammingForParallelComputing