Hewlett Packard Enterprise (HPE) has introduced a new small company supercomputer storage array called Cluster Storage for entry-level and mid-tier high-performance computing (HPC) and artificial intelligence (AI) computing clusters. This new ClusterStor line, acquired by HPE when they bought Cray in 2019, features a parallel architecture using solid-state drives (SSD) and hard disk drives (HDD) with Lustre file system software. The model, known as E1000, can scale from 60TB to tens of petabytes across multiple racks, each with up to 6.8PB of capacity. It delivers up to 1.6 TBps and 50 million IOPS per rack. HPE positions ClusterStor as storage for exascale, preexascale, and national AI supercomputers running Cray EX supercomputers.
The new Cray Storage Systems C500, based on the E1000, is designed for customers running modeling, simulation, and AI workloads on smaller compute clusters, often utilizing Cray XD systems. The XD665 mid-range system supports various networking protocols and provides direct connections between high-speed fabric, GPUs, NVMe drives, and CPUs. The C500 runs the same Lustre software as the E1000 but in a less expensive and more compact design. This system features a ProLiant DL325 server, metadata unit, scalable storage unit, and support for expansion chassis to increase usable file system capacity.
In terms of performance, the C500 offers between 22TB and 513TB of usable capacity from 24 NVMe SSDs, with up to 80GBps aggregate read and 60GBps write performance on the compute nodes. HPE is implementing software enhancements to make their storage systems easier to deploy and manage. However, specific details regarding these enhancements are hidden behind a restricted access HPE QuickSpecs web page.
Ulrich Plechschmidt, HPE product marketing for HPC and AI parallel storage, claims that the new Cray Storage Systems C500 will provide premium storage technologies at a fraction of the entry price compared to the E1000, with greater ease of use. The C500 is aimed at customers using smaller compute clusters for modeling, simulation, and AI workloads. HPE stresses the importance of their open-source-based Lustre support, ensuring that critical data resides in a system owned by a community rather than a single company.
Overall, HPE’s introduction of the Cluster Storage C500 aims to cater to the growing demand for high-performance storage solutions in the HPC and AI computing space. With a focus on scalability, performance, and cost-effectiveness, the C500 seeks to address the needs of customers working with smaller compute clusters looking to leverage the power of supercomputing technologies.
Article Source
https://blocksandfiles.com/2024/05/02/hpe-mini-storage-hpc-ai-training/