Programming Google App Engine with Python

Chapter 16. Task Queues and Scheduled Tasks

The App Engine architecture is well suited for handling web requests, small amounts of work that run in a stateless environment with the intent of returning a response to the user as fast as possible. But many web applications have other kinds of work that need to get done, work that doesn’t fit in the fast response model. Instead of doing the work while the user waits, it’s often acceptable to record what work needs to get done, respond to the user right away, then do the work later, within seconds or minutes. The ability to make this trade-off is especially useful with scalable web applications that use a read-optimized datastore, because updating an element of data may require several related but time-consuming updates, and it may not be essential for those updates to happen right away.

Consider a simple example. You’re the leader of your guild, and you’re scheduling a raid on the Lair of the Basilisk for next Thursday. When you post the invitation, two things need to happen: the raid is recorded on the guild calendar, and an invitation is sent to each member of the guild by email. If the app tries to do all of this in a single request handler, you might be staring at the browser’s spinning wheel for a while as the app calls the Mail API for each member of the guild. For a particularly large guild, the request handler might not even finish within 60 seconds, you’ll see a server error, and only some members will see the invitation. ...