China Underground > Science in China: News & features > Supermicro Advances AI Training with Cutting-Edge Storage Solutions

Supermicro Advances AI Training with Cutting-Edge Storage Solutions

Last Updated on 2024/01/30

Supermicro’s Advanced AI Storage Technology.

Supermicro, Inc. (NASDAQ: SMCI), a leading manufacturer of IT solutions for AI, Cloud, Storage, and 5G/Edge applications, has announced the introduction of a comprehensive, rack scale storage solution designed specifically to enhance data pipelines for AI training and inference. This innovative technology aims to streamline the process from data collection to high-performance data delivery, crucial for machine learning (ML) and AI operations.

Enhancing AI Data Pipelines

The new solution from Supermicro is tailored to maximize AI time-to-value, ensuring that GPU data pipelines remain optimally saturated. It enables the collection, transformation, and loading of extensive data volumes, reaching petascale capacities, into AI workflow pipelines. This multi-tiered approach has demonstrated its capability in delivering multi-petabyte data for AI Operations (AIOps) and ML Operations (MLOps) in real-world production environments.

Technical Specifications and Benefits

Supermicro’s solution boasts 20 petabytes (PB) per rack of high-performance flash storage. It is compatible with both air-cooled and liquid-cooled NVIDIA HGX H100 8-GPU based servers. This setup ensures accelerated AI and ML application processing at rack scale. The system can deliver a minimum of 270 GB/s of read throughput and 3.9 million Input/Output Operations Per Second (IOPS) per storage cluster. It can also be scaled up to handle several hundred petabytes of data. The integration of PCIe 5.0 and E3.S storage devices, along with the WEKA Data Platform software, results in a substantial performance boost for AI applications.

Comprehensive Storage Portfolio

Supermicro’s storage solutions are inclusive of Intel and AMD-based servers, crucial for AI data processing. The portfolio includes the Supermicro Petascale All-Flash storage servers, offering up to 983.04* terabytes (TB) per server and the Supermicro SuperServer with a 90 drive bay for larger capacity needs. These solutions are globally available and cater to ML, General AI (GenAI), and other complex computational workloads.

Solution Components

The storage solution is comprised of several key components:

  • All-Flash tier: Supermicro Petascale Storage Servers
  • Application tier: Supermicro 8U GPU Servers (AS -8125GS-TNHR and SYS-821GE-TNHR)
  • Object tier: Supermicro 90 drive bay 4U SuperStorage Server with Quantum ActiveScale object storage
  • Software: WEKA Data Platform and Quantum ActiveScale object storage
  • Network Switches: Supermicro InfiniBand and Ethernet Switches

Post Author

Previous

Taiwan’s Gogoro Launches Advanced ‘Pulse’ Smartscooter, Featuring Innovative Technology

Evergrande’s Crisis and Its Ripple Effects Across China’s Economic Landscape

Next

Enjoyed this post? Never miss out on future posts by following us

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.