How does Google App Engine's load-balancer work?

Question

I'm planning to build a web service on GAE that I anticipate having to deal with significant amounts of load. I therefore want to make full use of memcache such that hopefully, most requests won't need to go to the datastore.

There will be several different users of this web service, and each one is likely to require different, and not-overlapping data from the datastore. My hope is that GAE's load balancing will tend to send requests from the same user to the same Google App Engine instance, which will increase the memcache hit-rate.

Is this how the load balancer works?

Answer 1

Wether or not that is how the scheduling algorithms work, it is memcached job to show the same data to all instances, so if you cache your data on instance-A it will be available from instance-B.

This is the difference between memcached and just storing data in your app's memory.