HPE is pitching high performance compute customers with hybrid configurations of its Cray Supercomputing Storage Systems built for faster throughput, along with a Rack Management Controller that monitors cooling systems.
The hardware giant already supplies its Cray Supercomputing Storage Systems E2000 in all-flash, all-disk, tiered and hybrid configurations, but - we're told - the latest combination offers up to 41 percent more data throughput per storage rack.
The E2000 line embeds the Lustre open-source parallel file system, and is generally intended to fit NVMe Gen 5 solid-state drive (SSD) building blocks.
For the refreshed hybrid configurations, HPE is combining SSDs with new 32 TB SAS-4 (24 Gb/s SAS) HDDs and new enclosures supporting 74 or 100 drives.
HPE says it'll deliver up to 116 GB/s of read throughput and 92 GB/s of write throughput per rack, with up to 32 percent more usable storage capacity per rack for expansion to support the growing size of datasets in HPC, especially where sites are mixing AI and traditional simulation-style workloads.
One of the reasons behind these new hybrid configurations could be the ballooning cost of storage, caused by the relentless demand for AI infrastructure skewing the price of components such as memory and many other things. But prices for silicon-based NAND flash SSDs are growing faster than those of good old hard drives, as our colleagues at reported.
HPE says that as demand for data-intensive workloads grows, prices for large SSDs have soared. The company claims the new hybrid configurations let research institutions and national laboratories spend less on storage and put that money toward compute node purchases, without giving up on the performance they need.
“When making long-term investments in environments where advanced workloads are rapidly evolving, our customers want scalable, dense, and flexible systems to help them address dynamic shifts in requirements," said Trish Damkroger, HPE senior veep for HPC and AI Infrastructure Solutions.
The latest hybrid configurations are expected to go on sale in early 2027.
HPE's other introduction is a Rack Management Controller, providing what HPE calls "intelligent liquid cooling” via real-time monitoring, leak detection and automated control features. It integrates with HPE Supercomputing Management Software to gather telemetry and provide control.
Several platform-specific versions will be offered, each tailored to different rack-scale systems. Support will launch with HPE's own Cray Supercomputing GX5000 infrastructure, naturally, with versions for Nvidia’s NVL72 and AMD’s Helion and Helios kit to follow.
The Rack Management Controller will be available for HPE’s kit in March 2027, with support for AMD and Nvidia expected to follow in June. ®