Interest

I'm very interested in this, too. I have in mind a project that could use Amazon S3 to store gigabytes of data for each user, Amazon SimpleDB to store, index, and query structured metadata on users' data, and a Drupal-based Web site to allow users and administrators to access, search, and manage the whole thing.

While using S3 in this model seems relatively straightforward, the SimpleDB model is problematical because each item has to be retrieved in a separate HTTP request. Doing this serially a number of times to build a page would be hopelessly slow. I envision a separate process on the Web server that would accept requests from Drupal, issue parallel SimpleDB requests, synchronize on the responses, load them as a unit into memcached, and finally return the data to the caller.

Hope this helps to stimulate discussion.