[spicyj]

At Khan Academy, we actually have terabytes of data; some of our tables have close to a billion rows.

Our backend does much more than just serve static files -- you'd think that we don't have to do much but we keep track of video and exercise progress, summarize progress by user and by coach, provide video recommendations, host a Q&A system with notification support, award badges in real-time while avoiding costly datastore queries, handle mailing lists and email subscriptions, and have support for having a public profile to show off your progress, among other things.

We handle hundreds of requests per second and it's really nice to not have to think about scaling very much -- with any other solution, we'd be doing a lot more system administration.

[danmux]

Wow, my bad! Thanks for pointing that out. If you scale my numbers up that would be costing you $60,000 per month. If you are doing recommendations you are probably processing significant amounts of data as well. I guess theres no way we could know the commercial arrangements, but It surely cant be the same as your average small time developer.

[dylanvee]

We also use Hive on Amazon Elastic MapReduce to do some of our analytics.

[theodpHN]

Is KA billed standard public rates for GAE usage, or is this donated by Google?