Why is the number of "billed" instances so much greater than the number of "active" instances?

Question

Performing load tests on my app, I noticed that the Instances dashboard graph shows a pretty big difference between the number of active and billed instances:

What do active and total mean?

Also, after spending the day running load tests, here's what I see:

In the first peak, the number of billed instances pretty much matches the number of total instances. Then, on subsequent loads, the bumber of billed instances sits in between total and active.

Update 2013-02-21: I did another batch of load tests today, and I'm still seeing variance in where the billed instances stand relative to total and _active:

How are these numbers calculated? How should interpret them, considering that I'm trying to forecast our operational costs based on these numbers?

Answer 1

It seems (I believe) that if you have F2 instance in application settings each F2 active instance is counted as 2 billing instances. If you set F4 instances it counted as 4 billing instances. And so forth.

Total instances is number of instantiated but not billed instances - kind of "gift" from Google. If there would be more requests that need more instances GAE would not need to start a new instance but would use 1 from those "non-active". When the load is raising GAE start new instances but when the load is going down GAE would keep instances for a while but would not charge you for them. But they would be shut down eventually if load did not raise back.