Assistant Game Cloud

With the development of the national economy and the formulation of the national strategy of "rejuvenating the country through science and education", the country's investment in education is increasing. People are also increasingly aware of the role of HPC (High Performance Computing) High-performance computing technology in the development of the national economy and the country's comprehensive strength, as well as the huge potential and irreplaceable advantages of colleges and universities in developing HPC technology and carrying out HPC applications.

The outbreak of the epidemic in 2020 has brought changes to people's lives. We have added the concept of "cloud" to our subconscious mind, and "moved" what we usually do offline to online, which has also improved the High-performance computing market.

1. HPC – High-performance computing FatTwin with Omni Path

The HPC RA utilizes Supermicro's high-density FatTwin server and Supermicro's Omni Path switch in a fat tree topology. This design can achieve Petaflops' computing power and, together with a large capacity repository, can undertake the most demanding HPC tasks. The high-quality FatTwin system provides high-density computing performance, storage capacity, and power efficiency.

计算能力：

System: SYS-F619P2-RT (4U 8Nodes)
CPU: Dual Intel Xeon scalable processors, up to 56 cores/node (3584 cores/42U rack)
Memory: up to 1.5TB ECC 3DS LRDIMM, 1.5TB ECC RDIMM, DDR4 up to 2666MHz
Ultimate scalability: 64 computing nodes per 42U rack, 48 1U independent switches can support up to 1536 computing nodes!

联网：

Switch: Intel 100G 48 port Omni Path TOR switch with management card. High exchange capacity,
The total structural bandwidth is 9.6 Tb/s
L 2/3 switch: 1/10 Gb Ethernet super switch, 48 x 1 Gbps and 4 x SPF+10 Gbps Ethernet ports
Cost effectiveness: The FatTwin architecture makes every HPC cluster highly cost-effective!

冷却：

RDHx (rear door heat exchanger) cooler doors enable efficient cooling, with a cooling capacity of up to 75kW per rack

2. 参考架构– TwinPro加上Dragonfly / InfiniBand

该设计在蜻蜓拓扑中利用了TwinPro的速度和效率以及Mellanox Infiniband网络。蜻蜓拓扑结构可最大程度地减小网络直径并最大程度地增加带宽，从而可以在其顶部构建快速而强大的MPI系统。

计算能力：

系统：SYS-2028TP-HTR（2U 4个节点，总共64个节点/ 42U机架）
CPU：Intel Xeon处理器E5-2600 v4 / v3家族，每个节点最多44核（每个42U机架2816核）
内存：最高2TB†ECC 3DS LRDIMM，512GB ECC RDIMM和64 GB DDR4 LRDIMM-2666。

联网：

交换机：Mellanox InfiniBand FDR / EDR / HDR，1个U交换机，36个QSFP28端口
极高的带宽：Dragonfly架构与InfiniBand交换机一起，通过消除拥塞热点，确保了低延迟和最大的有效结构带宽！

冷却：

提供高效的DCLC（直接接触液体冷却）选件，每个机架的冷却能力超过35kW

High computing performance

Rich computing resources meet various needs of CAE applications and provide a suitable environment. It also uses the Intel Cryptography Hardware acceleration technology to reduce the impact of cryptographic computing on performance and ensure efficient use of resources.

High communication capability

This scheme uses InfiniBand with high bandwidth and low latency as the communication network to enhance communication capabilities between nodes, greatly reducing communication latency. And all computing nodes are fully interconnected through IB switches.

high reliability

The whole cluster adopts multiple reliability design schemes to greatly improve the reliability of the system and ensure that there is no Single point of failure in the system.