Software Faculty of Huazhong University of Science and Technology is one of the 35 national demonstrative software schools approved by the Ministry of Education. It engages in cooperative schooling with many famous domestic and foreign enterprises and provides education and training according to demands of the software industry. It meets international standards in teaching methods and develops software talents according to international software standards. It explores reforms and innovations continuously on schooling mode, management mechanism, curriculum and teaching content and method and relies on its superior teaching resources to develop numerous software talents for China’s IT industry.
The animation industry is a hot industry today with strong demands for and a severe undersupply of human resources. In order to satisfy teaching needs, Huazhong University of Science and Technology needed to establish a render farm urgently for teaching practices, operational guides and render management of faculties and students.
The rendering system platform of the Software School of Huazhong University of Science and Technology needs to satisfy the rendering requirements of high-quality big pictures compatible for movie requirements; the rendering speed needs to be 10 to 20 times higher than that of a single node; the sequence frame generated by rendering can be automatically stored where designated via the control by the render cluster management software; a high-performance workstation is also needed to satisfy the requirements on animation adjustment and modification.
Analysis of Application Features
The initial goal of the system is to satisfy the School’s application needs and determine the development principles of the rendering platform system with reservations for future expansions:
1. Isomeric: the rendering platform shall satisfy the needs for 3D modeling and rendering by school teaching. Therefore, the isomeric requirements have to be met during the development.
2. Open: the platform shall be an open one furnished with standard products and technologies where various professional rendering software can run on it.
3. Scalable: animation game technologies are updated rapidly. The cluster rendering platform shall lay a solid foundation for future expansion. With the School's increasing requirements on rendering ability, the render cluster shall satisfy the needs for expansion in the future.
4. Manageable: The maintenance and management work of the cluster rendering platform will be an arduous task upon completion. Therefore, the cluster rendering platform shall be easily manageable to make it convenient for subsequent management and maintenance work.
The system comprises 1 graphic workstation, 10 rendering nodes, 1 management node, 1 rendering distribution node and 1 IO node. The graphic work group comprises 38 nodes at the teacher graphic workstation units and the student graphic workstation units and it is mainly liable for 2D, 3D and video editing and other teaching work.
The rendering nodes are furnished with Intel Xeon processor for rendering computation.
The management nodes are liable for the installation, deployment and management of the cluster system and can achieve the collective monitoring and management of the whole cluster and hence accord substantial convenience to applications. The administrator in cluster operations is assisted by the 24-port 4-in-1 KVM switch.
The monitoring distribution node can realize the distribution and dispatching of rendering tasks and management of rendering work via the rendering management software.
IO node storage array is liable for the data sharing and storage of the general applications. It not only provides a highly reliable solution and redundant protection but also offers sufficient room for future system upgrading and expansion. The internal computing network is furnished with H3C core switch S5100 and 72*1Gb Ethernet interfaces for internal network data exchange. Ethernet interfaces are reserved for expansion in the future.
The rendering farm comprises graphic workstation and rendering system. Graphic workstation is used for 2D and 3D modeling. During the night, it is used as a rendering node to complete the rendering application together with the blades. Inspur Tiansuo TS10000 rendering node utilizes the most energy-efficient processing chip platform and the Inspur blade system has a power conversion ratio of 93% to save power for the client to the maximum extent. The Tiansuo monitoring and management software can monitor and manage Tiansuo TS10000 system collectively, remotely and visually across platforms.
When Inspur learnt about this project, there was only less than 1 month from the bidding date while its main competitors had followed this project for more than 1 year. Inspur raced against time to communicate with the client and had found some defects in the tender documents, i.e. the configurations could not fully satisfy the application needs of the school. To resolve this problem, Inspur communicated with the teachers, from the perspective of the client, and provided practical data to support its recommendations as well as helped the teachers locate and solve the problems. Considering the lack of rendering cluster management experiences of the school, Inspur proposed a full set of solution. As commented by Teacher Qiu of Huazhong University of Science and Technology, Inspur is a real solution provider rather than a mere software and hardware seller.