Cloudian Scalable AI Infrastructure with Reference Architecture based on Supermicro Servers and NVIDIA Accelerated Computing

Cloudian, Inc. announced a reference architecture for GenAI workloads that integrates Cloudian HyperStore object storage with Supermicro, Inc. servers based on NVIDIA OVX and NVIDIA HGX.

The solution has been shown to deliver throughput over 20GB/s per node, with performance scaling linearly to meet the demanding I/O requirements of modern AI workloads including model training, inference, and checkpointing.

Object storage is rapidly becoming a cornerstone technology for AI due to its ability to efficiently manage the massive volumes of unstructured data that power modern machine learning and deep learning models. The company delivers full S3 API compatibility, robust data protection, and scalable-performance enabled by integration with NVIDIA GPUDirect Storage, all of which empower organizations to unlock the full potential of AI innovation.

Reference architecture showcases several capabilities:

• High-performance flash, leveraging NVIDIA GPUDirect Storage and RDMA networking with linear throughput scaling
• Comprehensive enterprise feature set, including observability, self-healing and repair, multi-tenancy, QoS, dynamic storage provisioning, and data lifecycle management
• Enterprise-grade security certifications with built-in ransomware protection and no custom OS kernel

“The AI storage landscape demands extreme performance, exabyte scalability, and enterprise-grade data management capabilities,” said Amit Rawlani, Sr. director, alliances, Cloudian. “This solution revolutionizes AI infrastructure by delivering HPC-class performance with the scalability and cost-effectiveness of S3-compatible object storage, all while providing the management features that enterprises require.”

“This architecture combines Supermicro’s GPU-optimized systems with Cloudian’s enterprise object storage to create a complete solution for AI infrastructure,” said Lawrence Lam, VP, architecture AI and storage technology, Supermicro. “Organizations can deploy with confidence knowing the platform will scale seamlessly to meet their evolving needs.”

The reference architecture delivers native S3 API compatibility, security, and scalability, supporting configurations from entry-level deployments to massive enterprise AI implementations. This flexibility enables organizations to begin with modest installations and seamlessly expand their infrastructure as AI initiatives grow, while maintaining consistent performance and unified management across the entire environment. The architecture’s proven linear scaling ensures predictable growth without compromising system efficiency or administrative simplicity.

Resource:
Access the reference architecture document (registration required)