Northwestern Polytechnical University is located in Xi’an of Shaanxi Province. It is the only research-oriented, multidisciplinary and open technological university that focuses on aeronautic, astronautic and seafaring engineering education and science. With all-around development in science, engineering, management, art, economics and law, the University is also a key member of the National Project 985 and Project 211.
High-performance computing plays an important role in scientific research and many emerging fields are based on computation. This solution is designed to provide a stable, reliable and high-performance computing system for Northwestern Polytechnical University to satisfy its needs for super-large-scale computation.
Analysis of Application Features
The high-performance computer cluster of this project is mainly used for material computation, structure computation, fluid computation, electromagnetic computation, molecular dynamics computation and finite element computation in the aeronautic, astronautic, seafaring, material, E&M and other disciplines of the University. The main software includes Ansys, Fastran, Materials Studio, COMSOL Multiphysics, Abaqus, VASP, Gaussian and other commercial software, as well as Gamess, DUNS and other open-source software and self-developed programs.
Inspur has a professional HPC application and analysis team which uses unique test tools to establish the hardware platform requirements for client applications. Inspur is thus able to provide HPC solution with the highest performance-price ratio to its clients.
For this project, Inspur launched the Tiansuo TS10000 high-performance server cluster which is characterized by its outstanding computing performance, leading power control, convenient and easy monitoring and management system, open and scalable system and well-established service system. Moreover, the highly acclaimed Inspur high-performance cluster training service system is an additional asset to this project which further assists in alleviating the users’ concerns.
1. The system is furnished with the robust Intel Sandybridge E5-2670 processor and the CPU has a theoretical peak computing speed of 53TFLOPS; the system storage has a total capacity of 219TB, including parallel storage system of 179TB and optical fiber storage system of 40TB;
2. Inspur selects high-density blade servers to make up the computing server group which supports 8U20 node. It is more energy-efficient, more compact and easier for management, expansion and maintenance;
3. The system adopts Mellanox Infiniband high-speed network for its computing network featuring a fat tree structure to guarantee fully linear interconnection;
4. Latest Kepler K20 GPU acceleration node is configured to better assist the user to elevate the application level and capacity of the new architecture.
5. Inspur’s highly scalable parallel storage system TSExaStor has been adopted for the project. It has a higher bandwidth, higher stability, friendlier user interface and more backup strategies to offer the better user experience.
6. One IPMI network is configured for the convenience of the administrator’s management work outdoors.
7. Operation dispatching software Platform LSF, Inspur Tiansuo TSMM cluster management software, HPC application feature analytical software and high-performance computer service platform Cluster Engine are provided.
Total of 102 NX5440 blade computing nodes and seven blade cabinets I8000 are provided. Meanwhile, four GPU nodes, one Inspur TSExaStor parallel storage system and one Inspur AS1000G3 full-fiber storage system are also provided.
The system has as many as 160 computing nodes in total, including 153 blade computing nodes, 3 MIC accelerated computing nodes and 4 GPU accelerated computing nodes; L1 parallel storage server system includes 10 units with a total capacity of 179TB; one L2 storage system has a total capacity of 40TB.
A high-performance computing system is essential to the performance needs of scientific computing and its architecture needs to satisfy the application features. As explained by Teacher Gu of the Network Center of the Northwestern Polytechnical University, “Inspur fully understands the customer needs, and resolves the customer's problems from the perspective of customers and aims to satisfy the customer's needs in a real sense. Therefore, the cluster solution that Inspur offers to the University not only has leading performances and an advanced architecture, but also has high scalability for full protection of user’s prior investments.”
In order to maximize system performances, it is very important to optimize system software, application nodes, operating system and the network other than improving hardware performances. In these aspects, Inspur has provided timely, satisfactory and high-level services to its customers.