With a traditional HTTP server, the server is online 24/7, processing requests one by one as they come in. If the queue of incoming requests grows too large, some requests will time out. With Zappa, each request is given its own virtual HTTP "server" by Amazon API Gateway. AWS handles the horizontal scaling automatically, so no requests ever time out. Each request then calls your application from a memory cache in AWS Lambda and returns the response via Python's WSGI interface. After your app returns, the "server" dies.