Inspur Information built a converged platform for artificial intelligence (AI) and high-performance computing (HPC) to efficiently collect and analyze genomic data for a well-known research institute in South Korea
1. Jointly developed a massive data storage and analysis platform
Inspur has jointly developed dozens of PB data storage and analysis platforms for customers. Nodes are interconnected by a 200Gb/s InfiniBand network. The I/O read and write performance can reach more than 2GB/s, meeting the needs of the genomic information database, which includes massive data collection, high transmission, large storage, low latency, and high bandwidth.
2. Create diversified computing clusters and comprehensively improve application operating efficiency
Inspur has designed a number of business systems such as a biological big data storage system, a collection system, a high-speed analysis system, a sales management system, and an AI cloud service provision system. In terms of satisfying the high memory requirements of the biological big data high-speed analysis system, Inspur has adopted the eight-socket server fat node TS860M5 which can be configured with up to 12TB of memory in a single node. The excellent performance and extremely low physical space requirements minimizes the number of deployed nodes and significantly reduces TCO. For cloud-based AI model training and data processing, Inspur built a GPU cluster based on NVLink 3.0 with A100 GPUs and NF5488A5 AI servers capable of processing 5 petaFLOPS on a single node, which massively improves the training efficiency of AI models. It effectively supports the calculation of massive genomic data and helps customers efficiently deploy an AI development environment.
3. Provide a unified hardware management platform to reduce maintenance costs
Inspur has developed an automated operation and maintenance solution for customers by taking the Inspur physical infrastructure management (ISPIM) platform as the core and unified it with deployment, monitoring, operation, maintenance, and alarm management equipment from multiple vendors in the customer's data center. ISPIM's batch configuration function and out-of-band operating system deployment function can greatly improve equipment mounting efficiency; the 3D computer room function can completely restore the space and equipment layout of the data center. The parameters are clear at a glance, which improves the efficiency of fault prediction and improves overall operation and maintenance efficiency.
The genomic information database computing platform developed by Inspur Information for a research institute in South Korea meets the requirements of ultra-high-capacity genomic storage and computing, AI data processing, high-speed data transmission, and safe and efficient backup while also improving the operation and maintenance efficiency and helping reduce TCO by 16%. The South Korean research institute will standardize the storage and management of genomic data based on this platform, and provide extensive data foundation support for hospitals, enterprises, and research institutions to carry out genomic research. At the same time, with the integration of genomics and medical data, scientific research has been improved, the transformation of basic research results into clinical practice was accelerated, and the development of the genomic industry in the region was vigorously promoted.