The Research And Implementation On The Rendering System Based On Hadoop

With the fast development of computer and graphics technology,graphicsrendering technologies are widely used in3d simulation,virtual reality etc.Softwareand hardware technology of graphics update every day,however,single PC computerstill can not afford high realistic real-time rendering and efficient problem of complexscene.In order to improve the rendering speed,we need to efficiently manage therendering computing resources.In the context of this study, the paper focused on system architecture and taskload balancing issues.According to the characteristics of parallel rendering, we usedthe Hadoop distributed computing architecture,managing of hardware and softwareresources on the cluster to ensure the rational management of the rendering tasks,andimproving system utilization and throughput,thus efficiently improving the quality ofrendering services.Specific studies are as follows:First,base on the review of the status quo at home and abroad in parallelrendering,we introduced relevant theory and knowlodge of Hadoop and renderingtechnology,including the technical background of Hadoop,the key technologies inHadoop and parallel rendering technology.Second,in order to resolve the common load imbalance problem of single framesplit the task of rendering in sort-first architecture,we proposed a new dynamic loadbalancing strategy based on the kd-tree on the homogeneous with time feedback.Thealgorithm used inter-frame consistency,using rendering time of each rendering node tosplit the screen space,we can adjust the division of the rendering tasks according tothe changing of the scene in the real time,fianlly the rendering time of the renderingnodes tend consistent.The experimental data show that: this method can effectivelybalance the rendering load balancing among the nodes,especially to meet therendering requirements in the complex scenes.In addition,we put forward a kind of parallel rendering framework compatibletiming parallel and data parallel,which can achieve a switching between timingparallel and data parallel.Data parallel split single frame to task block,and renderingnode use OpenGL pipeline to concurrently render the corresponding sub-block,thusfurther enhance the rendering speed and interactivity.At last, A distributed parallel rendering system based on Hadoop was design edand implemented. the system can achieve the task load balance and reasonable scheduling,task parallel rendering,fault-tolerant operation on task, the real-timemonitoring the network and rendering nodes,the transaction processing of fault nodeand energy strategy and so on.the system has been successfully deploy and test inTianHe super calculate center in Defence Science University.The testing resultsshown that each module was running well, achived the desired effect and effectivellyrealized the automation parallel rendering pipeline,and partly solved the speedrendering problem.