Optimizing join query performance in google app engine

Scenario

• Entity1 (id,itmname)

• Entity2 (id,itmname,price)

• Entity3 (id,itmname,profit)

• profit and price are both IntegerProperty

I want to count all the item with price more then 500 and profit more then 10.

I know its join operation and is not supported by google. I tried my best to find out the way other then executing开发者_Go百科 queries separately and performing count but I didn't get anything.

The reason for not executing queries separately is query execution time. In each query I am getting more then 50000 records as result so it takes nearly 20 seconds in fetching records from first query.

Google App Engine developers have always focused on read optimization and that is one thing where denormalization pops in. Before you design your data structure, you should work on possible cases in which the data could be retrieved. Designing of models comes later. A closer look at I/O session about Building Scalable Web Applications with Google App Engine will prove helpful.

In the current situation, if you are interested in just the counts, you may go with a shard counter. It will require you to update every associated counter if the field updates.

Another approach involves performing nightly scheduled task which will do heavy calculations, and update counts and other stats you might need. You might find mapreduce helpful in this case. This approach will never give you real time data.

The standard solution to this problem is denormalization. Try storing a copy of price and profit in Entity1 and then you can answer your question with a single, simple query on Entity1.