DataDirect Networks (DDN), a Santa Clara, Calif.-based provider of data storage systems for high performance computing (HPC) environments, today announced the availability of the “world’s fastest” hybrid block storage system, the SFA14K, and its converged storage stablemate, the SFA14KE.

Scaling to nearly 7 petabytes (PB) in a single rack and delivering 6 million input/output operations per second (IOPS) and 60 gigabyte (GB) per second throughput in a 4U enclosure, the systems employ NVMe memory, SAS-based SSDs and traditional hard drives to speed up big analytics workloads. The SFA14KE adds compute capabilities, allowing organizations to run applications, virtual machines and files systems directly on the hardware’s embedded processor.

The SFA14K and SFA14KE can be outfitted with up to 72, 2.5-inch SAS or SATA SSDs or HDDs. Alternately, they can be ordered with up to 48 NVMe SSDs and 24 SSDs or HDDs. Additional 84-drive expansion enclosures supporting nearline SAS drives or SSDs are also available.

Compared to the competition, DDN claims the new flagships deliver 40x the performance. Versus its own hardware, the SFA14K and SFA14KE are 10 times faster than previous generations. Connectivity options include Fibre Channel (FC), InifiniBand, Ethernet and Intel’s Omni-Path HPC network fabric technology.

Contributing to the performance bump is the use of latency-reducing NVMe (NVM Express) SSDs that leverage the comparatively fast PCIe componentry to shuttle data to awaiting server processors more quickly than traditional storage interfaces.

Laura Shepard, director of vertical markets at DataDirect Networks, told InfoStor that NVME “is ideal for acceleration of non-bandwidth-optimized workflows – which these days is almost everything. Even traditionally sequential workflows have small IO components,” including metadata and small log files.

“NVMe attached SSDs for the first time, deliver enough performance for us to start looking at them to consider using as an alternative to in-memory computing,” continued Shepard. The technology proves particularly effective in processing “data that is too large to fit in system memory in both commercial and HPC” environments.

The technology can also help remove “large scale computing bottlenecks in traditional HPC,” she added. It can serve as a caching layer for metadata as well as a “layer for burst buffers,” she explained. “DDN IME [Infinite Memory Engine] takes advantage of NVMe to solve bursty workflow peak outputs, large-scale IO blender problems, places where performance bottlenecks are identifiable to a specific application or set of applications or a parallel file system.”