Interactive high-fidelity rendering in the Cloud

Abstract

High-fidelity rendering requires a substantial amount of computational power to render even moderately complex scenes. This is exacerbated by the need for interactivity, which can be only achieved on powerful high-end systems or computer clusters. Nowadays, cloud systems provide elasticity of computational systems and a seemingly infinite number of resources that can be made available to a client device. This work investigates the use of the cloud in providing interactive high-fidelity rendering to client devices irrespective of their computational power. A novel framework for rendering as a service is presented and evaluated. The framework supports scheduling policies for resource management amongst multiple connected clients; the policies are based on different job characteristics, from interactive jobs to batch-based online rendering. Results show that the framework is scalable and elastic, and incurs minimal realisation penalties; furthermore, it has been shown to sustain rendering of path traced images at 9 frames per second on a commodity hardware infrastructure.