A Novel Hierarchal Scheduling Strategy for Rendering System

摘要

When high performance computing platforms are widely used for rendering application, scheduling of rendering jobs becomes crucial due to the existing scheduling strategies are rarely balance fairness and performance. Moreover, the traditional task-centered assignment method cannot make full use of rendering resources. Aiming at these issues, a novel hierarchal scheduling strategy for cluster rendering system is proposed. The reservation based FirstFit (RF) job scheduling strategy and feedback based distribution for tasks (FD) strategy are together integrated to fulfill the maximization of system performance and load balancing. The experimental results demonstrate that the proposed strategy (RF-FD) outperforms in many cases than the existing scheduling strategies combining with the fixed-threads and naïve load balancing method while guaranteeing the fairness.