VentureBeat.com, March 26, 2018 by Dean Takahashi — Nyriad has built resilient storage systems that are accelerated using graphics processing units (GPUs), and today it is unveiling one of those systems in partnership with Advanced HPC.
These storage systems are built to handle such problems as the simultaneous removal or failure of 20 hard disks at the same time. Nyriad and Advanced HPC will show off the Orion storage systems at Nvidia’s GTC 2018 event in San Jose, California this week.
Last week, I ran into Nyriad’s chief technology officer, Alex St. John, a well-known techie who co-created DirectX graphics technology for Microsoft in the 1990s. He said, “Who would have thought that the GPU could be used to accelerate storage processing?”
The Cambridge, New Zealand-based company has raised more than $11 million to date for its Nsulate software for managing large-scale storage arrays in datacenters. The company is targeting the release of its first product in the first quarter of 2018.
Founded in 2014, Nyriad uses GPUs to converge computing and input/output tasks and minimize data movement during the processing of large data sets, significantly improving power consumption and accelerating performance for next-generation datacenters and supercomputers.
Nsulate is a GPU-accelerated block device that manages the storage array in Linux. The GPU does the resilience calculations and error correction for very large amounts of storage that operate at very fast speeds.
The companies say their new product achieves data protection levels well beyond any rival that uses RAID, or a redundant array of independent disks.
Advanced HPC’s Orion system is based on the Tyan Thunder SX FA100, which supports 100 hard disk drives in a 4-unit rack space, without the need for an external host unit. Orion is a first of its kind system that combines Nyriad’s Nsulate software with a Nvidia Tesla P4 and 64 gigabytes of Netlist NVDIMMs. It is targeted at high-performance computing, big data, and hyperscale storage in everything from large enterprises to life sciences.
It is being adopted by Oregon State University’s Center for Genome Research and Biocomputing (CGRB).
“For many years, the growth of disk capacity has far outpaced the growth of interface speeds, leading to extremely long rebuild times on large-scale RAID systems,” said Christopher Sullivan, assistant director for Biocomputing at CGRB, in a statement. “The CGRB at Oregon State University has played a pioneering role in introducing new hardware and software technologies to solve the growing problems around big data.”