SimpleDB vs Tokyo Cabinet

Has a开发者_如何学Cnybody compared SimpleDB and Tokyo Cabinet for performance and scalability? I'm coding my project against SimpleDB at the moment and considering benchmarking TC, be nice if somebody had already done it and could tell me whether it's worth testing my specific storage and searching operations. If not I'll run some direct comparisons and blog the results.

Project is using Ubuntu 9.1 & Ruby 1.8.7 on an Amazon EC2 small instance (for now).

I have done a project in TokyoCabinet, though several months ago, so the landscape may have changed since then.

I initially looked at SimpleDB because I was certain I was going to want to run everything on Amazon's EC2, but after actually looking at how much that would cost, it seemed pretty ridiculous, heh.

Also, the lag between my server and SimpleDB was unbearable. So on that note alone, I trashed the SimpleDB option and went with TC.

In actuality, I went with Tokyo Tyrant, and built a stupidly simple ORM on top of it that allowed me to read from the local TT instance and write to the "home" server, which then replicated out to all the remote servers (which ran in the cloud on demand).

That operation (writing "home" and replicating out) took extremely little time compared to SimpleDB. I have no idea why it was so slow for my server.

So in short, I went with neither, haha, but did stick to the Tokyo* line of software.