https://www.mdu.se/

The users of cloud services prioritize cost andperformance, but they increasingly demand sustainable practices.Sustainability is no longer a choice for businesses but a strategicimperative that shapes global industries. This paper presents anew system for utilizing the render farms in cloud data centers.The system aims to reduce energy consumption and costs in clouddata centers while maintaining a specific level of performance,particularly when rendering images and videos. The system canbe described as a cloud-based expert system that offers renderingas a service, while considering user preferences for performance,cost, and energy efficiency. The system reads different scenerendering parameters and accordingly chooses the most suitableGPUs that fit the user’s requirements. In other words, thesystem inputs are the scene complexity and user preferences.The output is the optimal GPU for rendering. Scene complexityis determined based on several parameters, such as the numberof frames and polygons, resulting in one scene-related value.The user preferences are also normalized to a preferences-relatedvalue. Then, these values are aggregated to determine the optimalavailable GPU to render the scene at the lowest cost, minimumpossible energy consumption, and highest performance.