I think it's unfortunate that GAE use the term instance-hour for billing, which is quite different from AWS instance-hour. GAE's instance-hour is process-instance-hour while AWS is machine-instance-hour.
The difference is quite important. In AWS, you can spin up dozens of processes in an instance. In GAE, the instance is the process. If the process is idle waiting for IO, you are still being charged. Spinning up more processes mean more instances running and more are charged.
In that sense, GAE's $0.08 per process-instance-hour is quite expensive compared to AWS's $0.085 per machine-instance-hour. "Single-thread" Python process needs multiple processes to handle multiple web requests which counted as multiple GAE instance-hours while only a single AWS instance-hour.
The difference is quite important. In AWS, you can spin up dozens of processes in an instance. In GAE, the instance is the process. If the process is idle waiting for IO, you are still being charged. Spinning up more processes mean more instances running and more are charged.
In that sense, GAE's $0.08 per process-instance-hour is quite expensive compared to AWS's $0.085 per machine-instance-hour. "Single-thread" Python process needs multiple processes to handle multiple web requests which counted as multiple GAE instance-hours while only a single AWS instance-hour.