Statistics reveal that China suffers annual losses amounting to 180 billion Yuan arising from climatic changes and climatic disasters. Research on climatic change thus becomes highly significant in China’s social, political and economic context. As such, Tsinghua University established the Earth System Scientific Research Center in 2009. The Computer Department, the Environment Department and the Nuclear Research Institute jointly established the research institute of global changes which is committed to researching on the frontier subjects of earth system science against a context of global climatic change and their socio-economic impact. It focuses on research on climatic system processes, earth system modes, earth system observations and global change economics and further carries out systematic scientific researches and develops talent for the issue of global climatic changes to promote the development the University’s earth system science.
The supercomputer platform “earth simulator system” of Tsinghua University is mainly used for numerical simulation of the natural process of the earth. It is a complicated system comprising high-performance computer, software tool, supportive technology, parallel application and application. It can recover or predict the natural change process of the earth and therefore is an important scientific research device for the human beings to know about the earth. With this system, the scientists can conduct simulative research on ocean currents, atmosphere and the earth’s crust. It forms the foundation for decision-making on major projects and major economic and political issues and plays a crucial role in disaster prevention, and reduction and prevention of climatic changes. The USA, Japan and other developed nations already started relevant work in the last century and have successfully forecasted the tropical storms along the Brazilian coast in 2004.
Analysis of Application Features
With increasing demands on HPC requirements by scientific research institutes and the increasing scale of applications, how to conserve user investments and reduce energy consumptions while improving system performances has become the focus in the development HPC technology today. Motivated by a sense of responsibility for users, Inspur have adopted many advanced technologies and launched the high-performance Tiansuo TS10000 cluster system based on the latest Intel Xeon processor with 32nm processing technique, with the aim of serving high-performance users with world-leading technologies and providing a highly efficient and stable scientific computational platform. The server based on Intel ®Xeon® processor can automatically and smartly balance performance and power consumption, and adapt to the workload and user demands on a real-time basis, thus delivering the best performance. Relying on the wealth of experiences accumulated in the HPC sector in the last couple of years, Inspur optimized configurations and provided Tiansuo TS10000 HPC cluster products to fully elevate the application efficacy and helped the user reduce expenditures.
The “Earth System Simulator”, a supercomputer with a capacity of over 100 trillion established by Inspur for Tsinghua University, is furnished with a super-isomeric parallel expandable architecture, super-high-computing density modularity, leveled aggregation high-speed network exchange, PB-level expandable massive storage system, self-adaptive smart liquid cooling system and other world leading innovative designs. The supercomputer with a capacity of over 100 trillion established by Inspur for Tsinghua University is advanced in the following ways:
High-performance: CPU-GPU synchronization acceleration technology is adopted; state-of-the-art Intel XeonX5670 six-core CPU is configured to achieve a peak CPU speed of 104 TFLOPS and a GPU accelerated computing speed of over 64 TFLOPS, with a total computing speed of 168 TFLOPS.
High density: Inspur blade server with the highest computing density in the industry is adopted; a single cabinet accommodates 200 latest six-core processors; a supercomputer system with a speed of 100 TFLOPS is realized in 9 cabinets only.
High throughput: advanced system I/O technology is adopted; QDR Infiniband network is configured to guarantee a communication bandwidth of 40Gb/S between computing nodes; a massive storage system with a bandwidth of 20GB/s and a total capacity of 1000TB is provided.
High energy efficiency: the industry-leading enclosed water-cooled cabinet technology is adopted to form an enclosed cold and heat circulation passage between the cabinet and the refrigerating unit LCP so as to fully satisfy the heat radiation requirements of the high-density cabinets with a power consumption of 35KW. The whole system is green and energy-efficient and all computing nodes are furnished with an energy-efficient power supply. The blade server has a power conversion efficiency of over 94% and the system peak power consumption does not exceed 350KW.
Tsinghua University’s supercomputer platform with a speed of 100 trillion is the first computing platform with a speed of over 100 trillion among China’s universities. It is expected to serve Tsinghua University’s trans-disciplinary research and provide computing and data services for China’s research on global change. This computer is responsible for the computation tasks involved in climatic simulation, forecast, and assessment and testing for the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC-AR5).