Hardware Configuration

无标题文档

The platform massive data processing and storage cluster uses a GPU + CPU heterogeneous system with a total of 3368 processors. It has 60 computing nodes, 14 fat nodes, 6 GPU nodes, and the total double-precision computing capacity of the cluster is about 371TFlops. One A 3.1P high-performance available capacity storage system is realized the measured aggregate read and write bandwidth of up to 40GB / s to meet applications such as huge data volume, high bandwidth, and high IOPS in the field of brain science. To ensure the efficient and stable operation of the cluster, the computing center also formulated The operation and maintenance management system, computer room inspection system, resource quota restriction strategy, computer room environment, hardware and software failure real-time alarm strategies were introduced.

Node class	queue/partition name	number of nodes	Processor Specifications	Number of cores per node	memory per node	public queue qos	Number of GPUs	Cluster computing power
Node configuration
cpu common node	q_cn	60	2*Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz	36 cores (Apply for a core, available 4900M memory)	192G	high	/	0.4PFlops
Four-way fat node	q_fat	2	4*Intel(R)Xeon(R) Gold 6140 CPU @ 2.30GHz	72 cores (Apply for a core, available 40960M	3072G	high	/
Incorporated into the public resource pool and the project team	q_fat_c	8	4*Intel(R) Xeon(R) Gold 6240 CPU @ 2.60GHz	72 cores (Apply for a core, available 20480M memory)	1536G	high/research group high_c	/
	q_fat_l	3	4*Intel(R) Xeon(R) Gold 6240 CPU @ 2.60GHz	72 cores (Apply for a core, available20480M memory)	1536G	high/research group high_l	/
	q_fat_z	1	4*Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz	56 cores (Apply for a core, available25600M memory)	1536G	high/research group high_z	/
GPU fat node	q_ai8	2	2*Intel(R) Xeon(R) Gold 5220 CPU @ 2.20GHz	36 cores (Apply for a core, available27648M memory)	1024G	high	8*NVIDIA Tesla V100 32GB
GPU node	q_ai4	4	2*Intel(R) Xeon(R) Gold 5220 CPU @ 2.20GHz	36 cores (Apply for a core, available14000M memory)	512G	high	4*NVIDIA Tesla V100 32GB
Note: Remarks: The default duration of the task is 2 days（There is no time limit for joining the research group of the public resource pool）. If you need to extend it for a longer time, you can contact the administrator to apply (there is no limit to the queue to which the research group queue incorporated into the public resource pool belongs) wangyanmmin@cibr.ac.cn
Storage configuration
Total available storage capacity: 3.1P, aggregate read and write bandwidth: 40GB/s

Cluster usage

Hardware Configuration