Chapter 17. Optimizing Service Calls

Handlers for user-facing requests spend most of their time calling App Engine services, such as the datastore or memcache. As such, making user-facing requests fast requires understanding how your application calls services, and applying techniques to optimize the heaviest uses of calls.

We’ve seen three optimization techniques already, but they’re worth reviewing:

  • Store heavily used results of datastore queries, URL fetches, and large computations in memcache. This exchanges expensive operations (even simple datastore gets) for fast calls to the memcache in the vast majority of cases, at the expense of a potentially stale view of the data in rare cases.

  • Defer work outside of the user-facing request by using task queues. When the work to prepare results for users occurs outside of the user-facing request, it’s easy to see how user-facing requests are dominated by requests to the datastore or memcache.

  • Use the datastore and memcache batch APIs when operating on many independent elements (when batch size limitations are not an issue). Every call to the service has remote procedure call overhead, so combining calls into batches saves overhead. It also reduces clock time spent on the call, because the services can perform the operations on the elements in parallel.

Another important optimization technique is to call services asynchronously. When you call a service asynchronously, the call returns immediately. Your request handler code can continue executing ...

Get Programming Google App Engine with Python now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.