Is it possible to dynamically allocate a Google Compute Engine Instance?

[Alex T. via StackOverflow]

I'm working on an application where I don't know beforehand how many instances of a Google Compute Engine I need, but theoretically the number of machines provisioned will have to scale in proportion to the size of my userbase.

So, my question is, is there an existing API for dynamically allocating and deallocating server instances on the fly? If so, is there scope for cloning an existing instance as a means to guarantee the availabilty of runtime dependencies, and is there scope for computing the billing cost of doing so in the Billing API?

If this isn't possible, are there any recommended design patterns for informing a server-delegation process of a newly available worker instance that can be deployed to?

Please DO NOT REPLY directly to this email but go to StackOverflow:
http://stackoverflow.com/questions/44015115/is-it-possible-to-dynamically-allocate-a-google-compute-engine-instance