I want to promote my small toy, which I wrote in Python, which uses App Engine as a platform, and in which I implemented the comet through regular polling.

Gameplay is a casual form of
risk-games . Everything is trite, take over the world and become a winner. So, get acquainted:
gumwars.com .
Tackle - details of the implementation and the calculation of the cost of support for App Engine.
Technical implementation
As you can see in the screenshot, the client side is implemented on the
Google Maps API , it is perfect for visualizing risk games. The server side is more complicated.
Although the App Engine can hold a connection for up to 30 seconds, only up to 30 connections can be processed at a time. Therefore,
long polling on App Engine disappears and ordinary polling remains. The solution is not effective, but initially I wanted to see how the App Engine will behave under such conditions, moreover, it is an excellent crash test. So, each user generates one request per second.
Calculating the cost of support and the relevance of the App Engine for polling
This whole undertaking is an experiment. And this experiment is not yet complete, I want to catch up with the audience and see how the platform will behave at a sufficient load. But still, a few predictable calculations.
1000 requests for receiving events fluctuate around $ 0.001-0.002, and since the game receives about 3,600 requests per hour from one client, we have a cost of about $ 0.0072 per person-hour of the game. Although, these calculations are quite conditional.
Plans
In the future I plan to write a small report, what came out of it. Also, later I want to hone the implementation of polling and publish a separate open source library (python and js parts).
Well, we are waiting for
WebSockets support for App Engine, of course.
')
Ps.
Unfortunately, there is no karma for the “I am PR” blog. Transferred to "I am promoting".