Load Balancing And Optimization Management Algorithm For GPU Cluster Cloud Rendering Platform

In recent years,3D scene rendering of virtual simulation has become a hot topic in computer graphics.In this thesis,the technology used in the current mainstream PC cluster cloud rendering platform is systematically summarized.At the same time,the advantages and disadvantages of cloud rendering platform are analyzed,and through some concrete examples at home and abroad,the transition from CPU rendering technology to GPU technology is introduced.It also introduces the responsible shading in cloud rendering The architecture of parallel rendering technology used in the main body GPU cluster is introduced.Then the programming tools used in this thesis and the CUDA and Optix ray tracing engine for ray tracing algorithm are introduced.This thesis focuses on the specific analysis and discussion of the GPU cluster load balancing technology,analyzes the significance of the load balancing technology by analogy,and puts forward the importance of solving the load distribution scheduling on the server from the reverse side of the load imbalance phenomenon.Because there are many different partitioning methods in the task scheduling strategy of cluster,this thesis explains the mainstream algorithm of load balancing scheduling from the typical scheduling strategy in detail.Finally,in order to demonstrate the feasibility and applicability of the cloud GPU cluster rendering capability and the proposed load balancing algorithm,a parallel rendering server-side GPU cluster rendering system is designed and implemented.Aiming at the problem of load balancing in cluster,the research on task scheduling strategy is carried out,and the analysis of load balancing strategy in the past is summarized.Then using the continuity in computer graphics to predict the time and the recursive depth in ray tracing,A load balancing algorithm for dynamic prediction of recursive depth is proposed.The algorithm firstly calculates the time of preframe rendering to the 3D scene.Initialization subdivision,and then according to the statistical value and the secondary light recursion in rendering,the subsequent rendering tasks are automatically dynamically partitioned.Therefore,the problem of cluster rendering performance shock caused by load imbalance is avoided.The experimental results show that the GPU cluster rendering system based on this load balancing strategy can render 3D real-time rendering performance.And efficiency gains,Therefore,the research on load balancing of GPU cluster cloud rendering platform in this thesis will promote the development of 3D rendering industry in the future.