Hi,

I have an application in .Net that needs to pre-compute a lot of information and process existing data.

I am looking for advice as to which cloud computing providers are best for the task, and also how many of them support .Net out of the box?
I've looked at Amazon but I'm not sure if it's available in my country.

The data set is about 50MB and my Core2Duo would need about 2 years to go through it all (I've optimized for a long time but still). I could make it less but the more CPU time I can buy the better.

Any recommendations and experiences would be appreciated, as I'm a novice in this area and it's quite unclear where to look as these are probably more professional needs.

Is your problem paralellizable? Can it even make use of many boxes?

.Net typically isn't used for that, something like Hadoop, which is Java based is more common.

With AWS you get as many boxes as you wish. Provide a disk image, it boots, and that's it. Windows is obviously available at slightly higher cost. It's probably the only provider which will be able to cater to your needs.

There is obvious Azure, but I never saw it mentioned with regard to compute tasks.

Heroku, AppEngine and similar are different kinds of hosting. For ad-hoc compute tasks, there simply aren't that many providers out there that I could recall. It's obviously different if you were looking for some supercomputing center, but those tend to have long waiting lines and don't run commodity technologies.

Yes it's paralellizable, and can make a use of many boxes.
It's pattern recognition so each instance would run with a slightly different range of values which it would handle in a reasonable time.

Thanks for your answer, I wasn't sure if I wasn't searching hard enough, or (like you say) that for "ad-hoc compute tasks, there simply aren't that many providers out there".

I'll probably try AWS again, it's just that last time I checked (1 year ago) they didn't accept my credit card because this country's not on their list and I had to go through someone else.
Maybe that's changed now, fingers crossed.

Hadoop has a streaming job interface, so you can use whatever will run on your boxes (we use python, you can use C# and mono if you wish).\

I think we also do something similar with CloudBase, but I am not familiar with that side of things.