A real-time game streaming optimization technique based on layer caching

Advances in cloud computing have enabled cloud gaming systems. In those systems, game logic and rendering is processed remotely in a cloud server and audio and video outputs are streamed to a thin client with limited computing capabilities, such as mobile phones, low-powered computers or even smart TVs. There are still many challenges involved in the task of providing cloud games, especially when trying to reduce server encoding time and video bitrate. In this work, we present a novel optimization technique based on image layers caching. Our technique allows that previously encoded layers are reused reducing encoding workload. In addition, only newly encoded layers are streamed, reducing the overall bitrate sent from server to client. We achieved around 23% stream size reduction with about 5% encoding time increase for cases where background cache usage is maximized.