Need Help?

BJTU Builds HPC Platform with Huawei Servers

2015/6/17 0:00:00


Beijing Jiaotong University (BJTU) is a national key university administered by China’s Ministry of Education and supported by the China Railway Corporation and Beijing Municipal Government. Among the first of China’s institutions to award master’s and doctoral degrees, BJTU was also one of the first universities selected for the National 211 Project and the 985 Strength Discipline Innovation Platform project. BJTU institutes are highly regarded in China, including the Communication and Transportation Institute, the Mechanical and Electronic Control Engineering Institute, the Electrical Engineering Institute, and the Electronic Information Engineering Institute.


As a leader in transportation and communication engineering in China, BJTU has introduced advanced research and teaching methods to improve scientific research and innovation. The university has assumed thousands of government tasks in addition to projects supported by the Ministries of Science and Technology, Education, and Railways, as well as other national scientific agencies. However, BJTU’s computing resources could not meet researchers’ requirements. The university needed a High-Performance Computing (HPC) platform.

HPC is widely used in physical and chemical research, oil exploration, biology, meteorology, environmental research, transportation, computer science, and engineering. Colleges and universities in China have begun to build scientific-research-orientated HPC centers, including Tsinghua University, Beijing University of Aeronautics and Astronautics, Beijing Normal University, Beijing University of Chemical Technology, and Beijing University of Technology.

To implement an HPC platform, BJTU needed components such as cluster management software, parallel environment software, compilers, and upper-layer application software. The university was also concerned that traditional SAN storage imposes high hardware and software costs, and might not meet HPC I/O and storage-capacity requirements.

BJTU also wanted to build an educational cloud computing platform. This platform would integrate the university’s IT infrastructure as unified resource pools and thus improve hardware utilization.


Huawei offers a complete technical solution for HPC. Huawei’s E9000 converged architecture blade servers configured with CH121 compute nodes implement high-density computing resources. A single E9000 subrack provides up to 16.5 Tera Floating-point Operations Per Second (TFLOPS), and the backplane supports 40 GE and 56-Gbit/s InfiniBand with a midplane capable of 32 Tbit/s switching capacity.

Huawei uses the RH2485 V2 four-socket rack servers to provide large storage capacity. One of these servers can incorporate as many as 48 memory DIMMs for a 1.5 TB memory capacity.

Huawei also uses RH2285 V2 servers to deploy General-Purpose computing on Graphics Processing Unit (GPGPU) nodes for graphics acceleration, floating-point calculations, and concurrent computing acceleration. 10 GE Huawei switches converge the 10 GE interfaces of all nodes.

Huawei’s HPC platform integrates the company’s FusionCluster cluster management software and includes other necessary software. FusionCluster provides powerful policy management by using a wide range of task and resource scheduling algorithms and supports user-defined scheduling algorithms. One-click functions handle configuration, installation, and service-based capacity expansion/reduction operations. FusionCluster can monitor and analyze multiple performance indicators and automatically take measures to save energy based on the results.

The Huawei HPC technology allows computing capability expansion to 100 TFLOPS without adding external 10 GE switches. Storage capacity can expand to the PB level using the Lustre open parallel/distributed storage scheme.


Huawei implemented the first phase of BJTU’s HPC platform in 2014. The platform provides 30.5 TFLOPS and implements unified resource planning and management. Huawei’s integrated technology enabled a rapid roll-out of the platform.

The HPC platform offers high scalability that meets requirements for system capacity expansion for years to come. Virtualization and cloud computing technologies reduce IT management and operating costs and improving hardware utilization for BJTU. The university now has the high-performance computing capabilities to support key R&D projects.