A trial on promoting data-intensive supercomputing through innovations in storage subsystem architecture

With the development and change of processor technology and computer architecture, high-performance computing technology has made considerable progress. However, the continuous expansion of data scale and the ever-increasing performance requirements of new types of applications such as big data and AI have caused traditional data-intensive computing to face huge data access performance problems. In this talk, we will summarize the data-intensive problems faced in our daily research, like high bandwidth and high IOPS requirements in climate application, multiple storage interface support issue in process of HPC log analysis, and introduce how we used OceanStor Pacific storage to overcome these issues. Finally, we will introduce some of our joint innovations with Huawei on HPC storage system for next-gen HPDA, e.g. IO optimization and storage-computing collaboration technology, Heterogeneous computing scheduling and offloading.