How many simultaneous clients can a Character Server service?
Character Server features such as dynamic Flash generation and question-answering can require considerable server-side resources. The total number of simultaneous clients that can be served by a single Character Server depends on many factors, but the gating factor tends to be the Text-to-Speech (TTS) engine. TTS requests are normally queued up, and the actual TTS generation occurs at several times real-time. For example, going from a text sentence to a flash stream consisting of a character speaking that sentence may require 100 milliseconds of intense CPU activity on a commodity server. During this time the audio is generated, compressed, and inserted into the Flash stream, along with animation frames selected from the character’s animation library. Additional requests are queued while a request is being processed, which will affect the latency, i.e. the total time between when a request is made and a response is received. In many cases the Server will improve the overall throughput
