Abstract

When constructing a high-performance computing cluster, one of the most pressing issues is ensuring efficient organization of the external storage systems. Such systems consist of both hardware and software components. The software component is a parallel file system, and consists of server and client parts. The article explores the principles of operating parallel file systems: mechanisms of data sharing between clients; metadata and data maintenance methods; as well as aspects of selecting a parallel file system for a high performance cluster computing cluster. We have had the opportunity to test some solutions on the large supercomputers of Joint Supercomputer Center of the Russian Academy of Sciences (JSCC RAS) and develop a basic model for file data flow in HPC cluster.

中文翻译：

---

高性能计算集群文件系统选型的期望

摘要

在构建高性能计算集群时，最紧迫的问题之一是如何保证外部存储系统的高效组织。此类系统由硬件和软件组件组成。软件组件是一个并行文件系统，由服务器和客户端部分组成。本文探讨了并行文件系统的操作原理：客户端之间的数据共享机制；元数据和数据维护方法；以及为高性能集群计算集群选择并行文件系统的方面。我们有机会在俄罗斯科学院联合超级计算机中心（JSCC RAS）的大型超级计算机上测试一些解决方案，并开发了 HPC 集群中文件数据流的基本模型。