Scientific research institutions · High performance computing

Project background

Scientific research computing generally faces challenges such as large amounts of data, high concurrent reading and writing, and many small files. Traditional storage cannot support the throughput and latency requirements of multi-disciplinary parallel computing.

Solutions

The storage node is built with GP5000 all-flash storage and paired with the LPFS parallel file system to provide a unified namespace and standard POSIX interface to support supercomputing and big data platforms.

• Single system up to 16 million IOPS, 72GB/s bandwidth
• Storage and computing separation architecture supports horizontal expansion
• Intelligent layering and life cycle management for scientific research data

Implementation value

It significantly shortens the read and write wait for scientific research tasks, improves the overall utilization of the cluster, and provides a stable base for scenarios such as meteorology, life sciences, and material simulation.