A Reliable Scheduling Algorithm for Render Cluster System

According to the problem in a rendering management system of time consumption in task scheduling and abnormalities in the procedure of rendering, the classical solution applies on the strategy of one-time scheduling. The strategy would increase the quantity of the frames which are not rendering successfully because of the accumulation of the frames which are not rendering, especially at the condition with a large number of frames in a scene. The anomaly probability of a render unit will be increased with the increasing quantity of slave nodes in a render unit. In this thesis, a new reliable task scheduling algorithm is proposed based on several dynamic scheduling for the purpose of decreasing render failure rate. The experiments are conducted to evaluate the performance of the proposed algorithms and compared with the one-time scheduling algorithm, the results that the proposed algorithm has a better performance in render failure rate and resource utilization.