This application has benchmarks for the three main libraries for 
GAE:
Pipeline--http://code.google.com/p/appengine-mapreduce
MapReduce--http://code.google.com/p/fantasm
Fantasm--http://code.google.com/p/appengine-pipeline

The benchmarks include:
Wordcount
Aggregation
Join
Subset
Grep

The code layout:
 - data
   # Contains code to generate data and delete it
 - wordcount
   # Counts the number of times each word comes up
 - join
   # Joins two tables based on a property field
 - grep
   # Search for a needle in a haystack
 - aggregate
   # Count up an integer on an entire entity kind
 - subset
   # Aggregate but only on a subset of entities