for me it's all about the cost (both time cost and money cost), and there is no comparison (if you can work with the GAE bigtable).
1. web2py does a good job of working with BigTable, so once you get out of the habit of joins, many apps will just work without much change in your coding. 2. ec2 really is just like a private server, you still have to install the server, configure it, create the machine image, deploy, monitor, update etc (time expensive) 3. GAE just runs. upload your app with a simple script (or press a button if you use the mac GAE launcher) and it just runs. no server to setup, monitor or deploy 4. GAE truely auto-scales. as more requests are made it just services them (until you reach your self-imposed budget limits) 5. AWS EC2 needs to be monitored for traffic and it is up to you setup and turn on additional servers (time and maybe money expensive) 6. AWS is paid by machine minute, GAE is paid by CPU cycles used. so always-on for AWS is at least $14 a month with a micro instance, GAE can serve 1000's of pages a day for free. 7. GAE background tasks is much harder than just writing scripts and cron. so if you need background services EC2 might be better. if you are curious, the iphone app starmaker (http://starmakerapp.com/) talks to a web2py GAE backend, but uses EC2 for some heavy audio processing (the website will migrate to web2py soon), and http://www.elizabethscanvas.org/ is complete web2py on GAE. that's my 3 cents. christian
