DCRA: Decentralized Cognitive Resource Allocation Model for Game as a Service

摘要

Game as a Service (GaaS) has rapidly emerged to the industry of cloud gaming. The power of GaaS lies on having one source of code with multiple users. Several systems were proposed to model GaaS. However, none has built a scalable and reliable model for such a service. The importance of having such a model lies on having an Internet-scale platform able to provide flexibility of different types of games genre and lower the barrier of end systems (i.e. mobile clients) while taking into consideration the probability of excessive loads and failures. In this paper, we implement a Distributed Cognitive Resource Allocation (DCRA) model to run mobile games on a large-scale distributed system. On the contrary of the existing centralized models, DCRA scales with the increase of mobile clients to handle high concurrent loads of clients' requests while providing a stable level of gaming experience. The results show that DCRA is able to scale well by providing almost fixed throughput and delay while increasing the clients requests load. Also, the system preserve its key features while simulating failures.