Some shortcomings that still irk me.

http://developers.slashdot.org/comments.pl?sid=329797&th...

Yes, it looks like Amazon has been solving the easy problems (paid AMIs, bigger instances) first, while the hard problems will be addressed later. But given that it took a year or so to add large instances, I wonder when load balancing and storage will be improved.

The 1.7TB drive on the extra-large instance seems strange to me. Who needs that much non-persistent storage?

Is the instance storage faster? I know there is discernible latency with file transfer between EC2 and S3. Having faster local storage for each instance could be very beneficial for swap space or temp tables, if your servers need a lot of it.

I still can't imagine needing that much temporary space and yet being small enough that using EC2 is still economical.

Some calculations don't need to be done that often. Recalculating for a massive data set can often be.. fudged for a period of time, in scientific computing - accuracy decreases over time, and then a full recalculation is eventually necessary. If you need major computing power, but only need it twice a month, this type of plan is economical, whether you're a large company or a small one.

The size of the enterprise is not always in proportion with the size of the datasets.

Obviously people won't use it as temp space; they'll use it to store databases and replicate them on multiple instances for reliability.

That still sounds unwise. One glitch slightly worse than the one from last week and you're badly screwed.

RenderMan does. The extra-large instance would be a good choice for rendering movie-quality 3D graphics.

Way to go! Really good news.