A well-known university in Istanbul, Turkey, is one of the most famous educational and scientific research institutions in the country. It is committed to engineering science and social science research. According to the 2016 QS World University Ranking, the university ranks first in the fields of "engineering and technology" and "natural science".
A well-known university in Turkey has the largest HPC research center in the country, which has been providing supercomputing and data storage services for academic and industrial users since 2006. The HPC Research Center mainly carries out theoretical research related to material physics and computational chemistry based on VASP, Gaussian and other commonly used research software for materials science and chemistry. Its high-performance infrastructure is upgraded and expanded yearly.
1. Excellent performance
The customer requires that the high-performance computing platform system provide excellent floating-point performance, with increased CPU performance, and strong cooperation with the GPU for excellent higher parallel computing capabilities.
2. Parallel computing network
Because material physics and computational chemistry software have high requirements for the parallelism, the university expects to build a cluster network using Ethernet and InfiniBand networks.
3. High system stability
The university expects to use a system that is easy to maintain and has high stability to ensure uninterrupted scientific research while efficiently processing data.
In order to further promote the scientific research capabilities of this Turkish university, and match the university's expectations for excellent performance of high-performance computing, high computing network parallelism, and high system stability, Inspur built a high-performance computing system based on its flagship server using the M6 platform. It supports VASP, Gaussian and other application software required in material physics research and computational chemistry research with Intel's new generation Xeon Scalable processor and Nvidia's new generation Ampere architecture accelerator, for maximum performance.
The theories of material physics and computational chemistry are complex, and their requirements for high-performance computing systems are also different. Inspur built a high-performance computing platform based on the 2U2S NF5280M6 server, which is capable of handling an array of applications and complex workloads. It has the advantages like high reliability and easy management that meet the university’s requirements for high-performance computing, high network bandwidth, and high stability. NF5280M6 is equipped with two Intel Xeon Scalable processor 8360Ys, which meet the university’s requirements for a floating point performance of 60TFLOPS for the cluster system and general computing of the processor, and flexibly adapt to the requirements of material physics and computational chemistry for complex computing power.
In addition, Inspur built a parallel computing network based on Nvidia's new generation HDR IB switch, provides the best performance network solution, offering a cost-effective interconnection solution and unparalleled high availability and reliability for high-performance computing clusters. It ensures 200Gbps communication bandwidth between all computing nodes, and achieves full line speed interconnection between all nodes.
The Inspur high-performance computing platform has become the cornerstone of scientific research projects in various fields for this Turkish university. The operation efficiency of VASP, Gaussian, and other application software required by the HPC Research Center in material physics research and computational chemistry research has been improved by more than twofold. Simultaneously, the stability of the high-performance computing platform based on Inspur servers has been improved, and the hardware failure rate has been reduced to less than 30%, which provides a strong guarantee for uninterrupted scientific research at the university.