Customers want to have multiple storage and file system choices to best address their workload needs. The HPC Ready Architecture for AI and Data Analytics offers multiple storage building blocks to meet customer performance needs based on their environment.
The architecture features:
The architecture also supports multiple parallel cluster file systems, which are designed to manage I/O-intensive workloads in performance-critical environments. One option is BeeGFS, which transparently spreads user data across multiple servers. File system performance and capacity can be scaled out by increasing the number of servers and disks in the system. Because high-performance I/O is a primary goal, the system has been designed as a high-speed scratch solution. At the core of the solution are high-speed NVMe SSDs that offer high bandwidth and low latency by removing the scheduler and queueing bottlenecks from the block layer. The BeeGFS file system also supports high aggregate I/O throughput.
The Dell EMC NFS Storage Solution (NSS) for HPC provides a scalable, high-availability NFS storage service that can be directly integrated into the system fabric. The core of the solution is an HA cluster, which provides a highly reliable and available storage service to HPC compute clusters by using a high-performance network connection such as InfiniBand or 10/25 Gb Ethernet (10/25 GbE).