HPE is to sell and support IBM’s venerable Spectrum Scale parallel file system for high performance computing (HPC) and AI workloads running on its ProLiant and Apollo servers,.
The company has added IBM Spectrum Scale to its HPE Parallel File System Storage offering, which is part of its HPC portfolio, and positioned under its ClusterStor array/ Lustre parallel file system. HPE already has resale agreements in place with Qumulo for scale-out filers and WekaIO for parallel access file software.
HPE’s Parallel File System Storage is said to support from tens to a few hundred compute node client systems, and scales out in terabytes. The high-end ClusterStor E1000 Lustre arrays support from several hundred to thousands of compute node clients, and scales out at the petabyte level.
The IBM tie-in was announced via a blog by Uli Plechschmidt who runs product marketing for HPC storage at HPE’s Cray business unit. “This really is a four-way winning scenario for all stakeholders – customers, channel partners, IBM, and HPE.”
He said that HPE Parallel File System Storage is unique, providing:
- The leading enterprise parallel file system according to Hyperion Research; IBM Spectrum Scale Erasure Code Edition (ECE)
- Running on HPE ProLIant and Apollo x86 enterprise servers
- Shipping fully integrated from HPE with HPE Operational Support Services for both hardware and software – one throat to choke
- Without the need to license storage capacity separately by terabyte or by storage drive (SSD and/or HDD)
NFS is Not For Speed
Rather than making smaller ClusterStor systems, HPE is reselling Spectrum Scale on its servers to take on DDN (GRIDScaler), Dell EMC PowerScale (Isilon), Lenovo (DSS-G), NetApp AFF arrays and Pure FlashBlade in the HPC and AI markets.
Pleschmidt characterises the above lineup – except for DDN – as NFS systems:.“NFS-based storage is great for classic enterprise file serving (e.g. home folders of employees on a shared file server). But when it comes to feeding modern CPU/GPU compute nodes with data at sufficient speeds to ensure a high utilisation of this expensive resource – then NFS no longer stands for Network File System but instead, it’s Not For Speed.”
VAST Data takes a different view, having shown it can deliver more than 140GB/sec to Nvidia DGX A100 GPU servers using ordinary NFS.
HPE has to take a different competitive positioning tack against DDN, which also sells Lustre-based systems. It says it offers end-to-end compute and storage support whereas HPE compute customers using DDN Lustre storage have two companies to deal with for support.
HPE Parallel File System Storage is generally available next month and can be used as a service through HPE’s GreenLake. An HPE white paper, Spend Less on HPC/AI Storage and More on CPU/GPU Compute, discusses HPE’s two HPC parallel file system offerings in more detail.