An autonomous resource provisioning framework for massively multiplayer online games in cloud environment

Abstract Massively Multiplayer Online Games (MMOGs) applications are a class of computationally intensive client-server multi-tier applications with real-time quality of service (QoS) requirements. In these online games, the players all over the world can interact with each other through a virtual environment. To guarantee the QoS requirements to a highly variable number of concurrent players, the game operators statically over-provision a large infrastructure capable of sustaining the game peak load, even though a large portion of the resources is unused most of the time and subsequently leads to incur high resource provisioning costs on MMOG providers. To address this problem, we introduce an autonomous resource-provisioning framework for MMOGs in a Cloud environment. Firstly, a load prediction service anticipates the future game-entity distribution from historical trace data using an adaptive neuro-fuzzy inference system (ANFIS) prediction model. Then, we proposed a fuzzy decision tree algorithm to estimate the proper number of resources to be allocated to each tier in the MMOG application using the predicted workload and user SLA. The experimental results under real and synthetic workloads indicate that the proposed solution outperforms in terms of accuracy and performance metrics compared with the other approaches.