Scale to Zero

[Jagath Weerasinghe]

Hello, 

I read that API Gateway supports scale to zero.

After it is scaled down to zero, how the API latency is kept minimal?

Does it support something like idle instances to reduce the latency of requests as in AppEngine? If not, what is the typical latency of a request when the API gateway is scaled down to zero? 

Thanks, 
Jagath Weerasinghe