German AI Research University Enhances Storage Performance with Xinnor and NEC Collaboration

One of Germany’s top universities for AI research has revolutionized its data infrastructure by deploying over 4PBs of NVMe flash storage.

This ambitious project was a collaborative effort between NEC Deutschland GmbH and Xinnor Ltd., leveraging Xinnor’s xiRAID software and the Lustre file system to meet the high demands of advanced AI workloads.

Facing a surge in data and increasing pressure on their existing storage setup, the university’s AI research center required a storage solution capable of handling the intensive computational needs of over 20 world-class ML research groups. The new storage system, built on NVMe SSDs and xiRAID’s RAID capabilities, ensures high performance, reliability, and scalability.

“Our xiRAID solution was specifically designed to handle the intense parallelism of NVMe devices,” said Davide Villa, Chief Revenue Officer at Xinnor. “NEC’s deep expertise in HPC and AI infrastructure has been instrumental in delivering this solution. Together, we’ve enabled a transformative leap in storage performance, ensuring the university’s researchers can focus on groundbreaking AI advancements, without worrying about data bottlenecks.”

Deployment includes dual-site infrastructure:

• First Location: 1.7PB of storage with NDR 200 connectivity, supporting 15xNvidia DGX H100 supercomputers.

• Second Location: 2.8PB of storage with 100Gb connectivity, supporting 28 nodes with 8xNvidia GeForce RTX 2080Ti each and 40 nodes with 8xNvidia A100 each.

The xiRAID solution, configured with RAID-6, ensures robust data protection and optimized performance for workloads involving small file operations. This configuration not only extends drive life but also minimizes downtime during hardware issues.

“NEC is proud to have partnered with Xinnor to deliver this cutting-edge solution,” Jürgen Waiblinger, manager storage solutions, NEC Deutschland. “Xinnor’s xiRAID software has proven to be a game-changer in NVMe-based storage. Its flexibility, reliability, and ease of integration with Lustre have provided the AI center with an infrastructure that not only meets today’s needs but is ready for tomorrow’s challenges.”

The project’s success underscores the growing importance of collaboration between technology providers and research institutions. With this deployment, the university’s AI research center has significantly reduced computation delays, enabling faster project turnarounds and increased research capacity.