The National Key Laboratory of Theoretical Chemistry Computation is the only national key laboratory for theoretical chemistry research in China. It constitutes the base for the development of outstanding theoretical chemistry professionals and the platform for foreign exchanges in theoretical chemistry. It has made major contributions to the development of theoretical chemistry in China. It passed the state acceptance in 1993 and formally commenced operation in 1994. The laboratory further passed national assessment in 1999 and 2004.
For a long time, the laboratory had relied on a small-scale system for high-performance computing. With the increasing computational load, however, the original small-scale units and servers could hardly satisfy the computational needs. With the development of cluster technologies, cluster computation became an inevitable trend.
Through repeated exchanges between Inspur’s high-performance computing engineers and the users, Inspur came to know that the cluster system would be mainly used for computational chemistry, life sciences and other relevant application fields. Improving teaching, scientific research and experimentation are essential to boost China's national strength in the scientific research on basic subjects and enhance the national scientific and technological competitiveness. The main applications of the lab include Gaussian, Material Studio, Gromacs and Namd.
Analysis of Application Features
1. High-performance computing platform: the system node unit system is furnished with the latest Intel 45nm four-core product with full-cache DDR2 internal memory and enterprise-level 15000rpm SAS hardware; the computing network is furnished with the advanced BNT core 10Gb switch to increase the system's parallel efficiency significantly; the storage system is furnished with high-performance FC-SATA disk array. The software platform is furnished with Intel mathematic library and encoding environment to ensure stable and reliable performances. The operation dispatching system is furnished with stable SGE.
2. Ease of use: the Tiansuo monitoring and management software developed by Inspur independently can monitor and manage Tiansuo T10000 system in collective, remote and visual manner across platforms. It can monitor the hardware information, software information, process information, user information and application information of various unit modules and perform real-time fault detection and various types of pre-alarms. Its robust mangement system allows resource module awakening, user management, database service management and Internet-based single-point configuration, hence increasing the management efficiency and guaranteeing high system availability.
3. High scalability: the system’s computing network can be easily expanded. Nearly 100 nodes can be expanded without adding network equipment. If network equipment is added, the system can be expanded to cover 200 nodes. The system expansion needs can thus be fully satisfied.
4. Low power consumption: Inspur has excellent performance in power consumption control, with a unit rated power consumption of 100W, which is even lower than that of industrial peers. This constitutes a key factor for the system solution.
In this project, the system mainly comprises 72 dual-way computing nodes and 18 four-way nodes, 3 management nodes, 3*IO nodes, 1*10Gb core network, 1*1Gb management network and 1 storage array. The system includes 896 CPU processing cores for calculation. When completed, the high-performance computing system can provide a peak computing capacity of 8.957 TFLOPS and hence offer the user affordable high-performance computing.
Management nodes are liable for the installation, deployment and management of the whole cluster system, as well as for the conduct of uniform monitoring, management and alarm of the whole cluster and accord convenience to the applications. The operation dispatching software SGE that runs at the management node facilitates task distribution, dispatching and management.
The storage system is designed on the basis of 4Gb optical fiber architecture. One FC-SATA disk array is adopted; the system has a storage capacity of 36TB for collective storage of data of various server nodes. The storage capacity can be expanded if necessary.
The cabinet and the control system comprise 8 Inspur Tiansuo high-performance cabinets and 1 KVM switch system for remote and collective control of the system.
Establishing a stable, reliable and easily-to-operate high-performance platform with cluster is the wish of all teachers of the School. Inspur has recommended the latest and best technology for the solution, which has indeed demonstrated better performances in terms of ease-of-use and power consumption. As such, Inspur ultimately won the recognition of all the teachers, including the advantages of 2U in heat radiation and scalability, the advantages of 10Gb in performance and ease of use, the advantages of management software in ease of use, and the advantages of operation dispatching system in reliability. Through long-term contact with the teachers, Inspur’s professionalism and dedication deeply touched the teachers who finally chose Inspur as the supplier.