Lenovo Group Senior Vice President and President of Greater China and Russia Chen Shaopeng said that the strategic cooperation between Lenovo and the Network Center of the Chinese Academy of Sciences will enable hundreds of scientific research institutes and institutions across the country to use the lowest cost** *Enjoy high-performance computing, services and storage resources. This breakthrough in high-end computing models will further improve my country's commercial computing capabilities and meet the "CEMS" needs of commercial customers at a higher level.
The "CEMS concept" is Lenovo's first concept for commercial technology development, that is, in the three major fields of public computing, enterprise computing and personal computing, it will continue to provide products and services of excellent quality to satisfy users. Needs for cost, efficiency, management and safety. According to Chen Shaopeng, under the current global economic situation, Lenovo strives to maximize IT efficiency and allow customers to achieve output for every investment. Lenovo has proposed the core concept of commercial technology - CEMS, which stands for Cost, Efficiency, Management, and Security. It not only represents customers' fundamental needs for commercial technology, but is also the source of sustainable development of Lenovo's commercial technology. With the continuous development of commercial computing, commercial technology with the "CEMS concept" as the core will create greater value for customers with more optimized operations and better products.
Lenovo Group Vice President Wei Jianglei said that the successful development of "Shenteng 7000" is an important step in building Lenovo's new generation of public and commercial computing platforms.
According to his introduction, the public and commercial computing platform adopts an innovative high-performance heterogeneous architecture and breaks through the traditional high-performance architecture through self-developed core technologies such as load balancing, integrated management, and automatic deployment. The application limitations of the computing platform include a variety of server nodes with different architectures to meet the diverse application needs of users. It can support both large-scale scientific computing and large-scale commercial computing. The successful development of "Shenzhen 7000" is An important step in building this platform.
Wei Jianglei also said that Lenovo will introduce professional operation and maintenance service teams through strategic cooperation with the government and large scientific research institutions such as the Network Center of the Chinese Academy of Sciences, and jointly invest to build a heterogeneous system company. The commercial computing platform provides scientific research and enterprise users with a low-cost, high-performance, easy-to-manage, and highly secure public and commercial computing platform.
Huang Xiangyang, director of the Network Center of the Chinese Academy of Sciences, said that the strategic cooperation with Lenovo Group is an important opportunity for the Network Center to accelerate its development in the tide of informatization. The Network Center will support the construction and development of scientific research informatization platforms. Development, carrying the important task of serving scientific research informatization and management informatization, and making active exploration and beneficial contributions to the improvement of national scientific and technological innovation capabilities and the innovative development of social economy.
This "strong alliance" between Lenovo Group and the Network Center of the Chinese Academy of Sciences is a further deepening and extension of cooperation between the two parties in the field of high-performance computing following the exascale high-performance computer project. Industry insiders believe that the "marriage" between Lenovo and the Network Center of the Chinese Academy of Sciences is a full manifestation of the integration of superior resources and the improvement of resource utilization efficiency in the field of information infrastructure services in my country, and is a breakthrough and innovation in high-end computing models.
The Shenteng 7000 Exascale machine adopts a hybrid high-performance computing cluster architecture, which consists of a cluster part using Intel Xeon processors and a fat node part using Intel Itanium2 processors. It is suitable for different Apply calculations. However, it should be noted that the program binary codes of these two parts of nodes are incompatible. The cluster part includes authentication server, front-end machine, management node, login node, startup node, I/O node, backup node and computing node.
Among them, the computing nodes include blades and thick nodes, and the configuration is as follows: 1140 ordinary (blade) nodes, each blade is equipped with two quad-core Xeon processors E5450, main frequency 3.00GHz, 32GB memory; 38 Thick nodes, IBM X3950M2 server, each thick node is configured with 16 quad-core Xeon processors X7350, main frequency 2.93GHz, 512GB memory; 12 visualization nodes (using rack-mounted servers), 2 Intel Xeon E5450, 32GB
Memory, Nvidia 9800GTS graphics card;
Other node configurations are the same as the computing blades, including 2 authentication servers, 2 front-end machines, 2
Management nodes, 8 login nodes, 12 startup nodes, 120 IO nodes, 2 backup nodes
; Operating system: Red Hat Enterprise Linux Server release 5.1, Linux kernel
2.6.18-53.el5.
The fat node uses two sets of SGI Altix4700 and uses Intel Itanium2 processor, which is backward compatible with Shenteng 6800 target code (can directly run some applications in the Shenteng 6800 environment) program, MPI program),
The configuration is as follows: The system consists of 2 computing nodes, 1 head node, and 2 file service nodes
◆ Fat node (computing node), node Adopting NUMA structure, equipped with 94 dual-socket blades
modules, 188 processors, 376 CPU cores, and about 2.5TB of shared memory.
Mainly used For batch job calculations;
◆ The head node is implemented by a fat node through logical partitions and is used for user login, compilation
and debugging programs, and job submission. The head node is configured with 4 dual-socket blades, ***16 CPU
cores, and 64GB of memory.
◆ The file service node is used by 1 SGI Altix450 through the logical partition function to become 2 CXFS file system servers. Users cannot directly access it. *** 384 Intel Itanium2 dual-core processor
◆ Model 9140M, 1.66GHz frequency, 666MHz bus,
◆ L1 Cache: 16KB (Data) 16KB (Instruction), L2 Cache: 256KB (Data)
1MB (instruction), L3 Cache: 18MB
◆ TDP 104W System floating point peak 5Tflops Total memory 5TB NUMALink interconnection is used in the interconnection network nodes (Bidirectional bandwidth 6.4GB/s, delay 1us); Gigabit Ethernet interconnection between nodes is used for management and NFS file system;
Two 20Gbps Infiniband interconnections are used between computing nodes.
The operating system adopts SUSE Linux 10SP2, and the kernel version is 2.6.16.60-0.21.default;
Others: System network: Gigabit Ethernet and 20Gbps Infiniband are used to implement interaction between nodes. Company; job management system, LSF HPC 7.0.
Global *** shared file system user $HOME directory: /home_soft/home uses SNFS file system; software installation directory: /home_soft/soft uses SNFS file system;
/work: uses SNFS file system, yes /datastore directory link; /luster: user workspace, using the Luster file system; /datastore/userdata directory: user important data storage space, using the SNFS file system, and data is regularly migrated to the tape library. /datastore/workspace directory: user’s temporary working directory, using SNFS file
system. Fat node dedicated workspace (visible only to fat nodes): /workspace uses SGI CXFS *** file system.
Thick node dedicated workspace (only thick node visible): /gpfs1, /gpfs2, using IBM GPFS
file system.