Cloudian, a long-time leader in scalable object storage technology, directly addresses the complexity of AI storage infrastructure with its new Cloudian HyperStore with NVIDIA GPUDirect for Object Storage.
Unveiled at the SC24 supercomputing conference in Atlanta, Cloudian now combines scale and performance in a single solution, promising to simplify the AI storage landscape with a unified data lake that can act as a shared data repository for all stages of the workflow.
Cloudian’s new HyperStore with NVIDIA GPUDirect for Object Storage is a significant step forward in high-performance data management. The seamless integration of GPUDirect with HyperStore allows users to benefit from Cloudian’s scalability and ease of use while enabling a dedicated high-performance data path for GPU workloads.
The Cloudian solution brings with it a compelling performance story. With GPUDirect, Cloudian reports continuous throughput speeds of 200GB/second from a HyperStore system. This is triple the throughput when using GPUDirect compared to traditional data access methods, as measured on otherwise identical all-flash systems. This high-performance throughput ensures that data is available to GPUs at the speed required for demanding workloads.
For AI workflows, this means faster insights, enhanced performance, lower cost, and streamlined operations. This is the first object storage platform to tackle the challenges of AI and performance GPU applications head-on, demonstrating Cloudian’s continuing drive to deliver innovation in object storage.
As NVIDIA GPUs become more central to advanced data-centric applications, Cloudian’s GPUDirect for Object Storage support is a compelling solution that will help enterprises across industries achieve more with their data while controlling costs and enhancing efficiency – a critical capability for enabling AI-driven digital transformation.
You can read our full Research Brief on Cloudian’s new solution at the link below.
