AI/ML has moved from experimentation with small data sets to a core strategy using as much data as possible to fine-tune inference and training models. As such, a thoughtful and thorough assessment of storage platforms has never been more critical.
According to a sweeping report from McKinsey & Co. in November of 2020, every organization team including HR, Sales, Finance and Supply Chain surveyed believed that AI and an intelligent approach to data storage benefited from larger and faster storage systems.
On March 23rd, HPCwire, NVIDIA and Pavilion will share how to accelerate GPU-based applications with NVIDIA validated Magnum IO GPUDirect using RDMA and RoCE to transform ordinary AI/ML experiences into extraordinary results.
It is setting up to be a big event where Pavilion and NVIDIA reset the thinking on parallel processing for GPUs, networking and data storage.
“Data sets are growing and more data needs to be analyzed in the same or shorter amount of time,” said VR Satish, CTO and Co-Founder of Pavilion. “The speed of business is only getting faster. Pavilion has redefined storage and its interconnect to deliver unrivaled choice and control for block, file and object data used in GPU-based applications and virtualized GPU environments.”
While Pavilion has taken the pole position in GPUDirect storage, there is a debate in the press about GPU and GPUDirect storage array performance metrics. Costa Hasapopoulos, Chief Field Technology Officer at Pavilion shrugs off the debate.
“Pavilion has the most performant, dense, scalable and flexible storage platform in the universe. Whether traditional CPU-to-GPU, GPUDirect or even Virtualized GPUs with VMware vSphere 7, we deliver unmatched performance and density across block, file and object data storage. This means storage is no longer a bottleneck for AI/ML pipelines.”
What can this mean for your organization?
How can you achieve amazing AI/ML results with a data storage rethink?
Tune in to the live event with HPCwire, NVIDIA and Pavilion on Tuesday, March 23, or register to receive a post-event link.
Register here.
Presenters:
