Speed of the different python-based HTTP-servers

[Thomas W]

I'm thinking of switching from my Twisted-HTTP-server-project to
CherryPy or Snakelets etc etc. The project involves alot of
database-access using SQLObject and SQLite ( perhaps MySQL and
PostgreSQL in the future ) and some heavy IO-stuff ( scanning of local
filesystem to extract meta information about images and storing this
info in the database. This won't happen too often though).

I don't expect this project to have alot of traffic once online, but it
would kinda suck if my software couldn't handle it if it really took
off.

So my question is; based on the very brief description above, are there
any of the python-based frameworks that fits my project better than the
others? And are there some of the frameworks that doesn't scale
particulary well? Benchmarks anyone?

My main focus is simplicity, one of the main reason I'm looking for
something else than Twisted. I'm a little conserned about the
threading-issues when using thread-based frameworks though, but if
speed and scalability would increase dramatically I'm inclined to
struggle with the threads anyhow.

Snakelets and CherryPy both look interesting compared to Twisted ( and
CherryPy seems to have some code available using SQLObject allready,
while Snakelets does not, so I'm leaning towards CherryPy at the moment
).

I want to use ZPT or simpleTAL for templating and SQLObject for
database-access etc if that makes any differences.

Thanks for any input you might have.

Best regards,
Thomas