Jeff, price is not the EC2's problem, its the offering.

It's Golem!

Ted Dziuba's analysis is, I'm afraid, full of FAIL.

Rather than relying on "top" of all things to deliver his CPU percentages, he should have actually benchmarked EC2 against a non-virtualized server. Like we have. And guess what? We get roughly the equivalent of a 1.7GHz Opteron on CPU-bound activities, not 50% of that.

The thing Ted apparently doesn't realize is that the physical EC2 hardware isn't exposed. It could be two CPUs, or four, or even eight. Then they use Xen to slice-and-dice it get it just right.

To the virtualized Linux environment, it can confusingly appear as if you're only getting 50% (or, say, 25% or 12% or something, in theory) of the CPU. That's because the physical box is much larger, but your instance only gets an appropriately sized portion of it.

Want more? They offer bigger sized instances that get larger/faster slices of the CPU (and everything else).

@onethumb: look at the threads in the AWS forums. people have benchmarked it. then kill yourself.

"dime of hour"

??

Can I haz that?

@macbeach: Do not want comment.

on second thought it looks like i fucked this up. my shit runs slower in ec2, i never benchmarked it myself.

meh. what ya gonna do.

My answer: [blogs.smugmug.com]

:)

@Ted Dziuba: Typical Persai-grade FAIL!

@onethumb: You fail worse than your site.