Deadline Sensitive Cloud Computing Resource Scheduling Method for Scene Rendering

Scene rendering is a key process in the area of modern digital animation and digital movie industry. It requires large amount of computing resource, as many digital movies usually contain thousands of scene frames and require many iterations of film editing. To provide flexible compute resources to the industry, cloud computing has become an effective platform for scene rendering. Previous works paid much attention on the load balancing problem of such platforms, which was motivated from service providers perspective. In this paper, we consider a new type of problem scene rendering with deadlines (SRD), which came from the users need. A simple way to meet the deadline is to provide exclusive computing resource for a given job. However, cloud computing providers pursue resource sharing among users. We proposed a method called deadline sensitive cloud resource scheduling (DSCRS) to solve the SRD problem, which supports both the objectives of resource sharing among scene rendering jobs and the satisfaction of job deadlines. Specifically, we model a SRD problem as a constraint satisfaction (CSP) problem, which is then solved by a state of the art constraint programming solver. And we proposed a new CSP encoding method that supports re-planning which is beyond the capability of the well-known flexible job-shop scheduling model. Preliminary results show that our method is effective.