Quantum Adds Support for NVIDIA GPUDirect Storage with Myriad All-Flash File System

Quantum Corporation announced development of a highly parallel file system client for the Myriad all-flash file system, designed to fully enable NVIDIA GPUDirect Storage capability while offering on-the-fly client node deployments with cross platform compatibility, including NVIDIA Grace Hopper with Grace ARM-based architectures.

This capability offers an approach to building AI/ML infrastructure intended to let customers add new GPU nodes to Myriad clusters as needed to respond quickly to evolving workflow and pipeline needs. Designed to maximize GPU utilization and performance, the client is optimized for GPU-intensive workloads such as AI/ML model training and inferencing, HPC visualization and modeling, and video rendering.

The Myriad parallel client design takes a groundbreaking approach to building out AI/ML infrastructure by installing directly on customer servers or workstations equipped with high-performance GPU cards, transforming their host workstation into operational Myriad GPU Nodes. By leveraging NVIDIA’s Magnum I/O GPUDirect Storage technology, the client establishes a direct RDMA data path between storage and GPU memory, bypassing CPU bottlenecks to enable exceptional performance.

Multiple client nodes can be added to a Myriad cluster at any time, and are a new connection option alongside SMB, NFS, and planned S3 access that can be added to any Myriad share point on-demand, giving customers the widest range of connection choice and flexibility available for all-flash file systems.

“Myriad’s parallel client development is guided by our vision to make Myriad the most capable, most flexible and easiest to use all-flash storage solution,” said Jeff Mulder, chief development officer, Quantum. “The new client brings unique capabilities by running as a fully integrated GPUDirect Node on client systems. Unlike traditional parallel file system clients, which rely on other appliances to perform these operations on their behalf, Myriad’s client performs its own metadata operations, data reduction, and data protection operations. This approach minimizes common bottlenecks, lets customers maximize their GPU investment across architectures, and allows performance to scale with the number of parallel clients accessing the Myriad system.”

The Myriad parallel file system client will be made available for evaluation through Quantum’s Early Access Program. This program is an integral part of the company’s product development process, enabling influential customers to test and validate new features and provide real-world workflow feedback that will benefit all users.

Quantum’s wide network of system integrators is ready to help integrate Myriad’s capabilities for mission critical content production needs such as animation/visual effects (VFX) rendering and AI/ML actions for rich media and video at scale.

“We are extremely impressed with the performance and capabilities of Myriad after testing it in our labs,” said Lance Hukill, chief commercial officer, CHESA, an integrator for media and entertainment customers. “The innovative approach to a parallel file system client will help our customers tackle the most demanding workloads that cannot be addressed with traditional legacy NAS storage systems and gives us tremendous flexibility in building customer solutions. We especially look forward to testing the new capabilities for large-scale high-resolution video rendering and development of new AI/ML tools to extract insights from large content lakes.”

Existing Myriad customers are eligible to participate in the Early Access Program for the new parallel file system client with NVIDIA GPUDirect Storage support upon its release in Q1 2025, with general availability in the 2nd half of 2025.

Comments

This is official now, it was unofficially announced during an IT Press Tour in March 2024, and it gives to Quantum a serious entrance in the AI game.

The key point here is double. First, Quantum is famous for its StorNext product and its client software to maximize the interaction with the back end storage. Second, at the same time, Myriad is a horizontally scalable file storage server software exposing NFS and SMB and soon S3.

The management team recognized a few things: it's difficult to address new workloads with StorNext, leader in the M&E domain, and a new product is needed especially to leverage new technologies such as NVMe, its network companion, container model, transactional KV store, erasure coding..., NFS and SMB are key for enterprise applications as they continue to be largely deployed and parallelism is paramount for high demanding applications such as AI/ML.

When you consider this, developing a client software to give parallel access to applications for Myriad make sense and Quantum knows how to do that. But what about pNFS, as it keeps NFS and offers parallelism at client level, why did not Quantum pick this path as it would have been a natural extension? It's the choice for Hammerspace, soon NetApp and potentially others.

Having this new client software gives to Myriad more potential market grip, becoming even more comprehensive accepting a mix of clients and access methods concurrently. As a storage platform, the idea is to offer the right access mode and expose different or same content whatever the client is or requires.

For a market perspective, it appears clearly that parallel access is a must have for file storage solutions and even if NFS is chosen, pNFS or specific client must be considered. Parallel file storage is here to stay.

More globally it confirms the software shift Quantum has started since Jamie Lerner arrived as CEO in July 2018.

We also notice that the Quantum stock, traded on the Nasdaq under the QMCO symbol, has a very impressive positive trajectory gaining 153% yesterday climbing to $60.02 coming from $2.98 November 19th.